This PDF file contains the front matter associated with SPIE Proceedings Volume 11427 including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee listing.

This paper presents a color recognition algorithm for natural scene images based on HSV color space.In this algorithm,different fuzzy membership are constructed on the hue channel,saturation channel and value channel.By establishing the fuzzy rule base,the Mamdani fuzzy inference method is selected,and the color images are divided into eleven kinds according to the human psychological perception of color.This algorithm can accurately identify the color of the image with rich color distribution and partial background color close to the target color. Experiments show that the algorithm's color recognition and segmentation results are in good agreement with human subjective visual perception, which effectively improves the color recognition accuracy of the robot.

Glass fiber reinforced plastics (GFRPs) are lightweight materials with high specific strength and high specific rigidity and is widely used in construction, machinery and other fields. With the improvement of economic levels and living standards, major cities in China have carried out a large number of water park construction. The major facilities in water parks are waterslides. The fastest waterslide in China is 25 m high at the running speed of 75 km/h. However, long-term exposure and ultraviolet radiation as well as the effects of alternating pressure of visitor weights make waterslides susceptible to thermal expansion and water seepage, which cause the surface damages to the glass fiber materials of waterslides. For detection of damages in GFRP materials, infrared thermography is more efficient, and more suitable than the conventional non-destructive technology and avoids coupling requirement, and electromagnetic interference [1] .

Simultaneous localization and mapping (SLAM) had been developed in the robotics community initially, and gradually applied to indoor positioning. The monocular visual SLAM lacks metric distances information due to scale ambiguity, so cannot be used in many application scenarios. This paper proposed a precision co-positioning and mapping algorithm based on WLAN and mono vision. This method can estimate the scale factor of monocular VSLAM and get the position in real scale space. Furthermore, the WLAN localization and monocular vision is fused by factor graph model to realize accurate positioning and sparse mapping. Experiment results show that the algorithm can estimate scale factor well and reached positioning accuracy of decimeter, which has good theoretical research and practical value in the field of mobile robot navigation in indoor and outdoor environment

Deep residual networks (ResNets) can learn deep feature representation from hyperspectral images (HSIs), and therefore have been widely used for HSI classification. Despite their high accuracies, there still exist a lot of challenging cases, such as open world recognition, limited-sample learning and visualization of learned classification features, which cannot be well addressed. Most of the challenges in HSI classification can be attributed to the dependence on softmax based loss function and classifiers, which cause the lack of robustness for deep learning models and the hardness to visualize the learned classification features. To improve the robustness and achieve the visualization of learned classification features, we propose a novel learning framework called Residual Prototype Learning Network, a combination of residual network and prototypes learning mechanism. Under the framework, a prototype learning based loss function is proposed to enhance intra-class compactness and the inter-class separation of these feature representations; in addition, a prototype learning based classifier is simultaneously proposed to achieve the 2D or 3D visualization of the classification features. The effectiveness of our proposed learning framework is evaluated on several publicly available HSI benchmarks, and the experimental results show that our approach achieve better results than traditional softmax based ResNets.

Based on the ultra-precision grinder, an experimental platform for single abrasive indentation was built to investigate the effect of single abrasive particle on the fused silica optical element in the grinding process. In the paper, the process of crack formation and propagation in the element was analyzed. Two crack growth nuclei were formed on both sides of the indentation point just after indentation. With the increase of the indentation depth, the crack size increased and those two crack growth nuclei joined together, and the longitudinal cracks torn and extended to the interior of the material. Experiments were carried out at different indentation depths, and the variation of force with indentation depth was obtained. The crack morphology was measured by microscopy, and the variation of transverse crack radius and longitudinal crack depth with indentation depth was obtained. The results had important engineering significance for improving the grinding process.

An imaging method of spherical antenna array is proposed by analogy with the imaging method of circular scanning synthetic aperture radar. After analyzing the principle of spherical antenna array, the feasibility of direct azimuth imaging using linear frequency modulation is discussed, and the azimuth resolution of spherical antenna array is calculated. The factors affecting the azimuth resolution of spherical array imaging, such as the ratio of spherical radius to wavelength, the number of array elements and radar operating frequency are analyzed, and the simulation results are given. At the same time, according to the principle of azimuth matched filtering, a weighted method is proposed to suppress sidelobes. Finally, the imaging effects of two kinds of element distributions on spherical surface with equal angle and concentric circle are compared.

In order to improve the stability of the removal function for aspheric surface and realizes its machining. A polishing tool based on non-Newtonian fluid (silly putty) is designed. The influence of shear stress of silly putty in polishing tools on the pressure distribution on the aspheric surface is analyzed. According Preston equation, the removal function of non- Newtonian fluid polishing tool is optimized theoretically based on its surface pressure distribution with the polishing tool executing a planetary motion of the polishing tool. The influence of the structure, materials and driving mode of polishing tool on the removal function shape was analyzed. Then, the polishing tool was improved and the marks on the removal spot were removed. The effect of filling rat io of silly putty and process parameters on removal function was studied by experiments. A method of obtaining stable removal function based on non-Newtonian fluid polishing tools is presented.

Rocket engine is a complex and accident-frequently-occurring system. Its fault location and cause analysis is challenge. This paper analyze a rocket launching fault based on the video captured by a visible light camera on the rocket during the flight. The primary mission is to judge whether the suspected "flame" is caused by the combustion luminescence or the reflection of sunlight by the fluids. A qualitative and semi-quantitative analysis is carried out from the perspective of intensity and color. The possibility of self-luminescence is analyzed, and the spectrum from visible to near infrared of hydrogen and oxygen combustion is measured on the ground. The atomic radiation spectra of Na, Ka and other related metal elements are detected. The combustion brightness and atmospheric scattering brightness are compared. Combining with other prior information of the engine, it is considered that the brightness of the re-burning of hydrogen and oxygen should be lower than the atmospheric scattering brightness. The real color of the suspected "flame" area in the image is analyzed by using the characteristics of the white balance processing algorithm. By comparing with other similar gray scale and similar color images, it is inferred that the real color is white, and the yellowish of the “flame” is caused by compensation of the blue background by the automatic white balance of the camera. Based on the above analysis, it is inferred that the "flame" is caused by the reflection of sunlight by liquid particles, not by combustion, and the region should be located at the bottom of the rocket.

Marine ship recognition is a challenging Fine-Grained Visual Categorization (FGVC) problem due to the large visual variations caused by motion blur, occlusion, lighting changes, and etc. The visual distinction between similar categories is usually very small, so it is difficult to solve it with general recognition algorithm. It demands an advanced discriminative model to accurately segment marine ships from the backgrounds and classify the type of the ship. However, effective models for the problem tend to be computationally prohibitive. To address these two conflicting challenges, we propose to recognize marine ship based on two cascade CNNs (convolutional neural networks), a shallow CNN and a deep CNN. The shallow CNN is used to quickly remove most of the background regions to reduce the computation cost, and the deep CNN is used to classify the type of ship in the remaining regions. The two CNNs are trained end-to-end, and they are complementary to each other to guarantee the recognition precision with low computation cost. Experimental results show that the proposed method is promising for marine ship recognition.

Researches have shown that using convolution neural network (CNN) on spatial-spectral domain can improve the performance of hyperspectral image (HSI) classification in recently years. However, due to the existence of spectral redundancy and the high dimensional kernels used in 3D-CNN, the HSI classification models are often heavy with a huge number of parameters and high computation complexity. Motivated by the lightweight model, this paper introduced a modular convolution structure named three-dimensional interleaved group convolution (3D-IGC). This structure contains two successive group convolutions with a channel shuffle operation between them. First group convolution extracts feature on spatial-spectral domain. Then the channel shuffle enables cross-group information interchange. After this, the second group convolution perform the point-wise convolution. We proved that an IGC is wider than a normal convolution in most cases by inferred formula. The empirical results demonstrate that the increment of width in 3D-IGC model is beneficial to HSI classification with the computation complexity preserved, especially when the model has fewer parameters. Compared with the normal convolution, the 3D-IGC can largely reduce the redundancy of convolution filters in channel domain, which greatly decreases the number of parameters and the computation cost without losing classification accuracy. We also considered the effects of the 3D-IGC on deep neural networks, therefore we used the 3D-IGC to modify the residual unit and get a lightweight model compared with ResNets.

Aiming at the improvement of edge effect in CNC polishing. A new polishing method based on the surface extension is proposed. The basic idea and workflow are presented. The availability of this method is verified by simulation. Experimental study was carried out on 420mm×420mm caliber fused quartz optical element. The experimental results show that this method can restrain the collapse and warped edge surface. PV less than λ/3，GRMS less than 7.7nm/cm and PSD1 less than 1.8nm can be obtained combining the small scale smoothing technology.

Aerial imagery target detection has been widely used in the military and economic fields. However, it still faces a variety of challenges. In this paper, we proposed several efficiency improvements based on YOLO v3 framework for getting a better small target detection precision. Firstly, a dual self-attention (DAN) block is embedded in Darknet-53’s ResNet units to refine the feature map adaptively. Furthermore, the deep semantic features are cascaded with the shallow outline features in a feedforward deconvolutional module to obtain context details of small targets. Finally, introducing online hard examples mining and combining Focal Loss to enhance the discriminating ability between classes. The experimental results on the VEDAI aerial dataset show that the proposed algorithm is significantly improved in accuracy compared to the original network and achieves better performance than two-stage algorithms.

Range-resolved laser reflective tomography is of great potential application in obtaining image information about an object with non-imaging laser radar system. The resulting time-dependent return signal which collected by non-imaging laser radar system provides one-dimensional raw projection data for reconstructing target image. However, this return signal can be regarded as the multi-convolution between the distribution function of target reflectivity, atmospheric transmission, detection circuit response and acquisition circuit response with the emitted laser pulse signal. In order to efficiently improve the reconstructed target image quality using short pulse laser, this paper presents a method used to restore the impulse response of target reflectivity modulation from the resulting time-dependent return signal so as to improve the reconstructed target image quality. This method is based on sending the laser and making it vertically irradiate to a profile of target in the front view to obtain basic wave pulse, then used it to recover the response of target reflectivity modulation. The experiment results show that this method is feasible and efficient.

In this paper, a research of PMMA optical fiber reflectivity changing along with the angle of tips has been studied. A type of integrated conical fiber probe is designed. Nine angles of cone fiber tips were fabricated and tested in 0.4 L /min continuous uniform bubble flow (air) in this study to validate the performance of each cone angle. A simulation model of PMMA optical fiber tip was established based on physical properties of fiber to simulate the reflectivity of different angle tips set into gas/liquid environment. The first and second high peaks of experimental data agree well with the positions of those calculated using the simulation.

GPS is widely applied in ocean measuring for its function of all-weather and real-time precise positioning. The theory of gauging currents and tides by using GPS was introduced and summarized the developments of various gauging methods, the advantages and disadvantages were discussed as well. With the rapid development of GPS satellite positioning technology, the inversion of ocean dynamic parameters has become a popular issue in recent years. Research on this area is still far from completing.

The conventional training for automobile assembly and disassembly has such weak points as need of ample space, lack of teachers, heavy loss of instruments, low efficiency and unsatisfactory training results. This article puts forward a kind of automobile assembly technology based on virtual reality (shortened as VR). With the help of HTC VIVE helmet display and handle as the input and output media, the interactive connection between the user and the virtual scene is established. Lead the structured automobile model into the Unity 3D development engine, and write the script using the object-oriented programming method of C# language, to realize automobile virtual assembly processes, which includes such major functions as main scene design, component recognition, assembly demonstration, virtual assembly and simulated driving. This technology, featuring rich content, short cycle, low cost, flexible and easy use, strong user experience, low risk, lead and play, and strong expansibility, etc., has good practical significance for current automobile assembly.

An engineering prototype of underwater wireless optical static monolithic communication system based on blue LED with high power and large field of view as communication light source and APD with high sensitivity and large target surface as receiving unit is established. The prototype adopts OOOK modulation mode, uses Fresnel lens with diameter of 75 mm and focal length of 25 mm as optical receiving antenna, installs narrowband filter, reduces signal-to-noise ratio, carries out communication coding through FPGA, and adopts staring direct detection mode to realize engineering integration of underwater wireless optical communication technology system, and intends to establish communication distance. The underwater large-field-of-view wireless optical communication link with a communication rate of 10 Mbit/s and a distance of 20 m. The performance parameters such as communication distance, power attenuation, communication delay, water disturbance and channel scattering of the system engineering prototype are measured and analyzed.

Coastal environmental elements such as bathymetry maps are of great significance to the economic and military development of each country. Spaceborne hyperspectral imager is one of the important instruments for coastal zone monitoring. Firstly, this paper systematically reviews the index system of spaceborne hyperspectral imagers, and then introduces the applications of hyperspectral remote sensing images in the retrieval of nearshore bathymetry. In order to improve the inversion accuracy, the current research status and shortcomings of fusion technology of laser active remote sensing and hyperspectral passive remote sensing are discussed. Furthermore, the index system of hyperspectral imagers is prospected based on the requirement of applications in coastal zone monitoring, which provides reference and support for the further development of hyperspectral remote sensing in coastal applications.

The space-dependent rate equations of passively Q-switched intracavity frequency-doubled diamond Raman lasers are deduced for the first time. The rate equations are normalized and solved numerically to investigate the influences of the normalized parameters (the normalized Raman gain coefficient, the normalized initial population inversion density, the normalized frequency doubling factor, saturable absorber parameter, the coupling rate of the pump beam and fundamental beam in the laser medium, the normalized loss factor, and etc) on the laser pulse parameters (pulse energy, peak power and pulse width). The new theories and numerical results will help design passively Q-switched intracavity frequency-doubled diamond Raman lasers of high performance.

Selecting a reliability matching area as template is one of the key issues to vision navigation. This paper proposes a metric for matching area selection based on line feature extraction and connection. Firstly, a new line feature is introduced to approximate the reliability information about matching area, which is called saliency line feature. Then, extracting method of these line features is put forward based on monogenic phase congruency model. Secondly, a convex shape descriptor is proposed to represent the spatial distribution characteristic of the line features by connection. Finally, a measure method is defined by merging the quantity and spatial distribution characteristic of the saliency line features, which can guide to select better matching area. The experimental results show that the proposed metric is valid and effective.

A novel short-term four-dimensional (4D) trajectory prediction model based on deep learning is proposed in this paper. The model is based on LSTM (Long Short-Term Memory) neural network. It consists of input layer, hidden layer and output layer. Original trajectory data is first preprocessed in order to form supervised learning sequences which are used as input of the model. LSTM cell is used in hidden layer, information flow from each LSTM unit to next moment includes the cell state and the hidden state, which can be used to implicitly model the motion state of the aircraft trajectory. Four-dimensional information of the predicted trajectory is obtained from the output of the model. Experimental results with real flight data show that the proposed method is more effective in improving the prediction accuracy and has better robustness to data sources than the existing aircraft performance models.

In the conventional single-image super-resolution algorithms, they assume that the sparse coefficients of the low-resolution patches and the corresponding high-resolution patches are the same. However, from our research, it is found that these coefficients are different most of the times. In this paper, the mapping relationship between the low-resolution coefficients and the high-resolution coefficients are learned based on neural networks. In this method, the low-resolution and high-resolution coefficients are first obtained from training images. Then, they are the inputs for a neural network to train this network. Finally, they are used in the reconstruction of the high-resolution image patches. Experimental results show that the proposed method has better performance than the original state-of-the-art algorithms.

Summer precipitation estimation is one of the key and difficult tasks in short-term climate prediction because of the large amount of convective precipitation in summer which is characterized by uneven distribution, large intensity, short duration and rapid change with time. In order to improve the accuracy of summer precipitation estimation, an efficient method by multi-time scale Support Vector Machine (SVM) with quantum optics inspired optimization (QOIO) is proposed in this paper. And the performance of the proposed method is verified by radar reflectivity and precipitation data of automatic weather stations (AWSs) in Shanghai. Using radar reflectivity and precipitat ion in the most relevant time scale, a rainfall estimation model based on multi-time scale SVM is established for each AWS to estimate next 6-minute precipitation. Compared with the traditional single Z-R relationship, linear regression, K-nearest neighbor and ordinary SVM, the results show the higher Threat Score and lower root mean square error can be obtained by the proposed method in summer precipitation estimation.

As an important maintenance measure, software reconfiguration is the key to detect the unreasonable part of the code module, namely code smell. Traditional detection methods rely on the experience of engineers, and the location efficiency of reconfiguration points is low. The existing automatic detection tools identify code smell with limited accuracy. Aiming at the problem that the number of reconstructed points in software system is huge and various, and the automation of reconstructed activities is low and difficult to optimize, the research framework of software smell prediction based on machine learning is studied and designed. Taking four common code smells as the research object, the classification algorithm and detection model of the best code smell are established, and the dimension reduction method of feature extraction is further improved. The highest accuracy rate is 89.8%, which can improve the automation level of software smell detection.

In order to study the tectonic characteristics and its relationship with uranium mineralization in Weijing area of Inner Mongolia, the remote sensing technique and visual interpretation are applied to interpret the linear structures in Weijing area with GF-2 remote sensing data, based on the atmospheric correction and another preprocessing. The distribution regularities of linear structures can be got from the field geological examination and the available geological data. The results indicates that the intersecting areas of NE-NEE trending structure F12 and NNE trending structure F11 with EW or NW trending structures are the favorable uranium metallogenic regions. The NE-NEE trending structure F12 and NNE trending structure F11 are the most important ore-controlling structures in the area.

Micro-machined gyroscope (MMG) has wide application prospects in the economic and military fields, especially in the military fields where weight and volume are important value. However, the precision of MMG are affected by the mechanical thermal noise, the Coriolis force of the driving mode coupled by the sensing mode, the quadrature error and the coupling damping. It is necessary to establish the more realistic gyroscope structural model and improve the closed loop driving control. In this paper, the structural model of MMG which can describe the mechanical thermal noise, the Coriolis force of the driving mode coupled by the sensing mode, the quadrature error and the coupling damping is proposed firstly. For the proposed gyroscope model, the closed-loop driving control based on the reinforcement learning PID algorithm is applied to improve the precision of MMG. The simulation results show that the reinforcement learning PID controller can fulfill the requirements of the rapid start-up and the overshoot reducing. The control system can stabilize the amplitude and track the resonant frequency of the driving mode. It is important for the applications of MMG.

An important problem for both image processing and computer version is to synthesize the novel view of a 3D object. We propose a shared conditional adversarial auto-encoder (SCAAE) network that is trained end-to-end on the task of rendering previously unseen object given a single image of this object. The model uses the advanced GAN framework to build the generator by introducing U-net, which can generate a novel view image based on the input image and a controllable condition signal. The FCN model is used to construct the D-network to distinguish real and fake images. We also propose a new objective function which considers both the distribution consistency and transformation persistence. We designed a SCAEE network to generate multi-view images of objects, instead of the three dimensional effect of physical models, which solves the shortcoming of artificial modeling. Experiments demonstrate that the new network structure is better than other already existing.

The spatial resolution of multi-spectral remote sensing data can satisfy multi-scale structure interpretation. The authors took Qinglong area in Hebei as the study area, utilized the comprehens ive advantages of ETM+ , Spot6 and WorldView3 data of the study area to interpret the geological structures at two different levels from ETM+ data with 15 m resolution and Spot6 data with 6 m resolution to WorldView3 data with 0.5m resolution, and obtained a goo d result. Firstly, based on ETM+ data interpretation structure, the band operation results of SPOT6 data are used to indirectly reflect the structure informat ion, and the high resolution WorldView3 data is integrated to construct the interpretation analysis. Finally, the interpretation structure information is revised based on field validation. The results show that multi-source remote sensing data have the characteristics of high resolution, accuracy and high efficiency in geological structure interpretation.

The rapid prediction of carbon content by spectral method is helpful to grasp the carbon dynamic of sediments in intertidal zone in a timely manner. The spectral reflectivity of sediments with high carbon content and low carbon content often varies widely, limiting the accuracy of the model. When sediment samples are classified according to the carbon content and modeled separately, accurate prediction of sediment carbon is expected to be achieved. However, previous studies have not reported it. In our study, sediment samples were divided into low-carbon and high-carbon sample sets according to 3 g/kg carbon content, and divided into low-carbon, medium-carbon and high-carbon sample sets according to 2 g/kg and 4 g/kg carbon content, and then were pre-processed and PLS modeled separately. It is found that these classification can improve the modeling accuracy, but not improve the prediction accuracy. These results can provide a technical reference for the prediction of sediment carbon in intertidal zone.

Visual object tracking is one of the most attractive issue in computer vision. Recently, deep neural network has been widely developed in object tracking and showing great accuracy. In general, the accuracy of tracking task decreases dramatically when the background becomes complex or occluded. Here, we propose an end-to-end lightweight Siamese convolution neural network to achieve fast and robust target tracking especially for infrared target. The network structure replaces the hand-crafted features by the multi-layers deep convolution features of the target, so that higher precision can be achieved. Specifically, object location is updated in every frame by refreshing a response-map. However, the success rate of tracking task decreases dramatically when the background becomes complex or occluded. Consequently, a simple and robust anti-occlusion tracking method is presented. The tracking accuracy is evaluated during tracking process by computing the tracking confidence parameters. The parameters are composed of two parts: target confusion degree which indicates the degree of background interference and target occlusion degree which indicates the degree of target occlusion. Once the target is occluded, the location of the target object is corrected immediately. Experimental results demonstrate that the proposed framework achieves state-of-the-art performance on the popular OTB50 and OTB100 benchmarks.

Begin with the most pressing general questions base on the major surveys of astronomy and astrophysics today, the paper introduce the mission of deep space infrared detection, and the main requirements for infrared detectors, then introduce the principles and performance of cutting edge infrared detectors, readout integrated circuit and Application Specific integrated Circuit (ASIC) for deep space observatory. Base on present state of the infrared detectors, propose advices for developing Chinese infrared Focal Plane Array (FPA) detector for deep space observatory.

To meet the requirements of automation and intelligence of the vehicle-sighting system, the paper gives an overview of two widely-used target recognition algorithms, including pattern recognition and deep learning. Four typical algorithms, HOG-cascading-Adaboost algorithm and surf combining SVM algorithm, which belong to pattern recognition, CNN network and YOLOv3 network, which belong to deep learning, are elaborated in detail. Different algorithms are used to identify images in the same test set in the experiment, and the performance of each algorithm is compared from three aspects, recognition rate, recall rate and recognition time. Finally, it can be concluded that YOLOv3 algorithm is better for target recognition when concerning the recognition rate and recall rate, with the recognition rate as high as 95.8% and fewer targets missed. Considering the real-time effect, the pattern recognition algorithm has less recognition time but the recognition rate reduces. Therefore, the recognition time and recognition rate should have a compromise in practical application.

In order to improve the accuracy of numerical weather prediction(NWP) temperature, a support vector machine (SVM) model based on LASSO feature analysis is proposed to revise the predicted temperature for the next 12 hours. In this paper, high-resolution mode prediction data that include 2m temperature and related meteorological factors forecasted by the European Center of Medium range Weather Forecast ( ECMWF) , and the temperature data of the automatic stations in East China and coastal areas provided by the Shanghai Meteorological Bureau are used to build the proposed model. , In this paper, The results show that the root mean square error, absolute error and accuracy are greatly improved by the proposed prediction model. The comprehensive performance of the proposed method is better than that of the traditional linear regression technology.

The environment perception and navigation ability of robots are the basis of robot interaction with environment. The scene understanding is the prerequisite of robot autonomous navigation and navigation is the ultimate goal of robot scene understanding, therefore, a mobile robot semantic map navigation system is designed. Firstly, a dense pointcloud map is constructed by using a deep convolutional neural network. Then, the mapping relationship between the depth camera data and the laser data is proposed, for purpose of realizing their interconversion. And a method of constructing two-dimensional raster map by simulating laser data by using the data of the depth camera is proposed. Finally, a method that semantic information is integrated into the two-dimensional grid map is proposed, realizing the semantic navigation of the service robot. The experimental results showed that compared with other robotic navigation systems, the proposed algorithm can better realize the interactions between service robots and people and the environment.

Great progresses have been achieved on object recognition in the wild. However, it still remains a challenging problem due to tiny objects. Likewise, post-disaster or post-war personnel search and rescue missions often encounter specific problems of complex and wide areas and weak human targets. This work proposes a weak human target recognition technique for unmanned personnel search and rescue. Firstly, referring to human visual cognitive mechanism based on visible light detection equipment, deep convolutional neural network is applied to enable the model to perceive the targets in complex background. Then a single-shot detector is designed to extract the scale adaptive depth feature to match the regions of interest. Last, a deep network model compression algorithm and an efficient online processing framework are designed to speed up and detect the targets.

This paper is based on the deep learning license plate recognition system, which is a method of deep learning in the recognition of license plates.In the recognition of license plates, improved Convolutional-Neural-Network (CNN) is used to identify the accuracy and speed of recognition. The experimental results show that the application of convolutional neural network in license plate recognition can effectively improve the recognition rate of the license plate in various environments such as pollution, insufficient illumination, etc. This recognition rate is improved by means of a large training character set. The more character forms included in the character set, the higher the recognition rate, the more the license plate character recognition rate can reach 98% or more. In addition, for the trained convolutional neural network, including the license plate extraction and pre-processing recognition speed can also reach less than 30 ms.

Two-step phase-shifting sectioning structured light illumination microscopy (TSSIM) that reconstructs a three-dimensional structure using Fourier transform is proposed. Undesirable background signals corresponding to out-of-focus signals are eliminated using this method. Since there is no restriction for accurate phase shifts, this method does not suffer from large retrieval errors, unlike conventional sectioning structured illumination wide-field fluorescence microscopy (SSIWM). It can be used directly without modifying the conventional SSIWM microscope setup employing two of the three captured images, and can be applied to both shape measurements and biological observation. Less exposure time is required; thus, photobleaching and phototoxicity in biological observation are reduced. Further, the impact of the phase-shift difference on the signal-to-noise ratio of reconstruction image is analyzed. Both simulations and experiments are presented to show the validity of the proposed method.

Imaging simulation becomes an effective and necessary way to see how the wide-field sensor performs because design and analysis of a wide-field sensor requires a lot of rigid conditions. Here, a long-wave infrared panoramic scanning sensor is designed for small target detection. Space-based imaging simulation based on the characteristics of the panoramic scanning sensor is realized. Serial images covering 360°×10° field of view is generated. Stars, long-distance satellites, and space debris appear as point sources in the images. It is demonstrated that the panoramic scanning sensor has the capability of observing a long-distance small target and can be used for space target detection.

Fringe projection profilometry is a mature 3D measurement technique, however it is still difficult to measure high dynamic range objects in real-time. In this paper, we present a method of using digital micro-mirror devices to dynamic range enhancement of fringe projection profilometry. After applying this method, we can obtain two fringe images with different exposures in one projection period. These two fringes will be combined to solve the high dynamic range problem. Furthermore, we also build a multi-view system to realize real-time measuring stably because the multi-view system can retrieve the absolute phase without projecting additional fringe patterns. Experiments demonstrate that the proposed method can realize real-time three-dimensional measurement at 28 frames per second and the dynamic range of fringe projection profilometry system is increased by 2 times.

Visible-near infrared reflection spectroscopy has the advantages of fast and green, and has great technical advantages in the field detection of soil components, but no study on organic carbon (OC) in marine beach sediments has been found. According to the OC content and visible - near infrared reflection spectrum of sediment samples from Jiaozhou Bay beach in Qingdao, the spectral preprocessing was carried out by S-G derivative filter, and the sediment samples were divided by Kennard-Stone algorithm. On the basis of the whole band, the prediction model of sediment OC is established by using the least square support vector machine (LSSVM) algorithm. According to the evaluation results of the model, the modeling set R²=0.97, the prediction set R²=0.83, and the relative analysis error RPD=2.46, shows that the model has a good prediction effect.

Based on LOS method ,combined with DSMC, the particle clouds movement model and the optical radiation transmission model were established, the dynamic radiation performance and transmittance of particle clouds in space within medium wave and long wave were analyzed accordingly, the results showed that the heat flux in space has a certain influence on the particle clouds radiation, but not on the extinction obviously; when the number density of particle clouds was greater than 1x10⁷ , the transmittance of particle clouds in medium wave and long wave was less than 10%.

In this paper, a temperature control method of flexible composite film was proposed based on porous adsorption principle and phase change materials blending technology, the flexible composite film samples were developed accordingly, and the temperature control performance was tested experimentally. It was shown by the test results, the phase change of flexible composite film can be applied to a wide temperature range of 23-55 ℃, and also conducted heat well.

The near-sensing detection system based on the software-defined radio (SDR) platform is widely used due to its versatility and algorithm reconfigurability. Frequency modulation continuous wave (FMCW) detection has high precision and strong anti-interference ability. To solve the problem where the near-sensing detection system has a large fixed error when the modulation frequency deviation is small, a ranging method based on spectral feature matching is proposed in this paper. Compared with traditional algorithms, such as harmonic detection, zero-padding Fast Fourier Transform (FFT) and beat frequency correction, the proposed algorithm makes full use of the spectral characteristics of the beat signal. By establishing a spectral feature library of different distances, a matching evaluation function is given to transform the problem of distance recognition into a feature matching problem. The accuracy of the algorithm is verified by comparing the ranging error of the feature matching algorithm and other ranging algorithms at different signal-noise ratio (SNR) levels.

Aiming at the low accuracy and low efficiency of manual defect detection of stamping aluminum sheet, a surface defect detection based on modified PHOG algorithm is proposed in this paper. PHOG is a spatial shape descriptor that introduces spatial pyramid model to achieve multi-scale representation of the target. However, in the field of target recognition with similar shapes and different details, the PHOG lacks a description of the relationship between pixels in the image. In order to improve the detail description ability of PHOG algorithm, the sliding window mechanism is introduced to improve PHOG. Based on the improved PHOG, combined with the advantages of SVM, the defect detection algorithm flow is given. The experimental results show that, the proposed defect detection algorithm has good accuracy and real-time performance.

In multi-agent systems, agents coordinate their behavior and work together to achieve a shared goal through collaboration. However, in multi-agent systems, selecting qualified participants to form effective collaboration communities is challenging. In this paper, we propose a minimum circle covering algorithm, as a solution for on-demand participant selection for collaboration in multi-agent systems. Furthermore, a twenty-one point FBG sensors are divided into four sensing function agent in Structural Health Monitoring (SHM) system is experimented in an aircraft wing box. Correspondingly, there are four intelligent evaluation agents and one system collaborative agent in the multi-agent intelligent health monitoring system. For the damage loading position prediction on the aircraft wing box, the collaborative participation selection strategy based on the minimum circle coverage is verified experimentally. The research result indicates that the minimum circle covering algorithm can be used to select the participation in multi-agent intelligent health monitoring system, of all the participations in the collaboration, it enables them to identify and select a qualified participants.

Vehicle recognition is a key issue of intelligent transportation systems. For traditional color-based vehicle recognition methods, an important problem is that color information of vehicle is susceptible on lumination variation, which will inevitably reduce the recognition accuracy. Another is that there are contrastive interfering vehicles with similar color in same exposure environment, which could result in misrecognition of target vehicle. This paper presents a novel technique for recognizing target vehicle in different lumination scenarios via spectral feature. This method consists of three parts, ambient light spectrum calibration, vehicle spectral reflectivity detection and vehicle recognition. And an prototype system named Selectable Imaging Spectrum Detection System (SISDS), has been setup by this spectral feature-based method. A new combination algorithm is proposed for the detection part which contains the deep learning model YOLOv3 and the discriminant tracking algorithm KCF. Also, a CNN model is designed for improving the recognition accuracy through transforming the spectral data into two-dimension matrix. The experimental results of the SISDS clearly demonstrate that the spectral feature-based method overcomes the shortcomings of the traditional color-based method, and it can not only distinguish the vehicles of different colors in the case of overexposure, but also can recognize the target vehicle from the same color interference vehicles under various lumination, the recognition accuracy is up to 96.4%. Compared with color based method, the spectral feature-based method has great lamination robustness, and high recognition accuracy.

In recent years, with the rapid development of laser and optical detection technology, as a kind of spectroscopic technique for the qualitative and quantitative analysis of elements, laser-induced breakdown spectroscopy (LIBS) has been more widely used in the field of deep space exploration. In this paper, the development history and significance of human deep space exploration are briefly introduced, then the basic principle and development status of LIBS are outlined, and then the latest application progress of LIBS in the field of deep space exploration is introduced in detail. Finally, the application prospect of LIBS in deep space exploration is prospected.

Multi-class objects detection in remote sensing image is attracting increasing attention recent years. In particular, the method based on deep learning has made outstanding achievements in the object detection. However, the deep network is easy to overfitting for the insufficient of remote sensing image dataset. What’s more, the current deep learning- based methods of object detection in remote sensing image usually ignore the context information of the objects. To cope with these problems, a novel object detection method based on regularized convolutional network and context information are proposed in this paper. A form of structured dropout method is used in convolutional layers to dropping continuous regions. To address the problem of lack of context, spatial recurrent neural networks are used to integrate the contextual information outside the region of interest. Comprehensive experiments in a public ten-class object detection data set show that the proposed object detection method has an outstanding detection accuracy under different scenarios.

We investigate the effects of moderate-to-strong oceanic turbulence on the propagation of orbital angular momentum (OAM) modes carried by Laguerre-Gaussian (LG) beams. The analytic expression of the signal OAM mode detection probability is derived based on the extended Rytov theory. Our results show that turbulence with smaller dissipation rate of mean-squared temperature, larger dissipation rate of turbulent kinetic energy per unit mass of fluid and smaller temperature-salinity balance parameter can improve the detection probability. The LG beam with a smaller azimuthal mode order, larger radial mode order and longer wavelength can be less affected by oceanic turbulence. Our results also indicate that Bessel-Gauss (BG) beams possess much better resistant ability to oceanic turbulence than LG beams in the moderate to strong turbulent ocean because of their non-diffraction and self-healing characteristics.

Coherent beam combining (CBC) is recently used to generate high power vortex beams which are strongly required in specific applications. In this paper, based on the extended HuygensFresnel principle, the exact theoretical forms for the intensity distribution of the CBC Bessel-Gaussian beams (BGBs) in turbulent ocean are derived. To show the superiority of CBC BGBs in turbulent channels, the comparison in the intensity evolution of the CBC BGBs with the ideal BGBs is performed. It is found that the beam spreading of the CBC BGBs is smaller than that of the ideal BGBs under the same oceanic turbulence conditions. Moreover, the effect of the different optical source parameter on the quality of the CBC BGBs is also analyzed. The results show that the increasing number of the beamlets causes the optical energy to be more concentrated, and thus leading to a longer non-diffraction propagation distance.

Low repetition-rate and long pulse-duration laser around 2 μm has been investigated and widely used in deep space exploration, thanks to its high single pulse energy, low atmospheric loss, and safety to human eyes. This paper briefly summarizes the investigations of kilohertz and nanosecond laser around 2 μm, in terms of three pulse modulation techniques. In particular, the discussion of the advantages and disadvantages of the three pulse modulation techniques can further promote the development of the deep space exploration in China.

To inspect the treatment effect of high power plasma on fused silica, inductively coupled plasma with power of 36kW at atmospheric pressure , CF₄ as working gas, was applied to carry out the etching experiments of fused silica, in which the effect of CF₄ flow rate on material remove rate, surface roughness and surface profile were investigated. The experimental results showed that the material remove mechanism of fused silica treated by high power inductively coupled plasma was a result of the comprehensive effect of melting and evaporation at high temperature, chemical etching and bombardment of high energy ions. The fused silica remove rate reached 0.208g/min with the CF₄ flow rate of 2.8L/min. The surfaces of fused silica samples became rougher with the increased CF₄ flow rate, as the subsurface micro-cracks were opened and formed etching pits by the plasma etching. The surface global profile was affected seriously by the distance between outer plasma flame and samples while the local profile was dominated by etching pits.

The principle and method of existing indoor shadow photography system and shadow target calibration system are introduced. The effects of muzzle shock wave, projectile shock wave, trajectory descent of projectile flight and natural wind, dust, ground irregularity and stray light in the field measurement environment when using Shadow Photography to measure the Motion Attitude of Large Caliber Projectile are analyzed. The key parameters of the large caliber projectile motion attitude measurement system in the field are determined, and the key links that need to be paid attention to in the design are put forward accordingly.

The optical-mechanical model of the off-axis optical system with oscillating mirror is established. The stray light path is analyzed on the basis of the simulation results. According to various types of stray light paths, the main path of stray light are determined and the influence of important surfaces and illumination surfaces on stray light of the camera are clarified. Effective restraint measures are proposed without blocking the imaging light. After re-simulation, the veiling glare indexes of the two imaging channels are greatly reduced. PST (Point Source Transmittance) is used as an index to test the validity of stray light suppression measures. When the off-axis angle is 30 degrees, the PST of the camera reaches 10-8, which meets the design requirements.

Vortex volume grating (VVG) has advantages of high diffraction efficiency (DE), excellent angular selectivity, broad spectral band, and flexible design, making it a good candidate as an orbital angular momentum (OAM) generator in high power laser system.. In our research, a vortex volume grating with a relative diffraction efficiency of 84.58% and an angular spectrum of 1.864 mrad at 1064 nm is fabricated, and the refractive index modulation (RIM) is retrieved.

The morphology evolution and optical characteristic within longitudinal direction of self-organized nanopatterns induced during Ar⁺ ion bombardment (IB) with Fe co-deposition on fused silica (SiO₂) surfaces are investigated in this work. At the same ion fluence and angle of incidence, simultaneous Fe co-deposition can result in significant terrace morphology effect with increased roughness and horizontal feature size compared to conventional pure IB. Due to the interaction between the pure IB and Fe co-deposition, well-ordered corrugation patterns can be produced on the surface of the fused silica. Such structure can be used to synthesize ordered self-organized metal nanowire arrays on dielectric materials. This method combines nano-patterning of ion bombardment, impurity co-deposition and metal shadow deposition effect, which provides a feasible way to replace traditional continuous nano-processing methods. Due to the excitation of metal plasmon resonance, nanowire arrays exhibit different optical transmission spectral characteristics at different position in the structure region and show optical anisotropy in the longitudinal direction due to linear changes of structure horizontal feature size.

Using the meteorological statistics from NCEP reanalysis data, the detection distances of point source targets at sea in medium and long waves are calculated. The difference of detection distance between different locations in two infrared bands is analyzed. And the detection distance difference of the same location in different months of the year. The results show that the influence of season on detection distance is greater in long wave than in medium wave. There is little difference in detection distance between different locations and months of medium wave. The detection distance of long wave varies greatly with location and month. Of the three sites surveyed, the largest difference was 20%.

In view of the previous work, the effective radius of curvature (ERC) and Rayleigh range z_R of a general-type partially coherent beam propagating are studied in anisotropic oceanic turbulence. Based on the extended Huygens–Fresnel integral, we derive the formulae of the second-order moments of Wigner distribution function (WDF) in anisotropic oceanic turbulence. The calculation results reveal that turbulent anisotropy has a sharp impact on both the ERC and Rayleigh range for a general-type beam. In the near field, the effect of anisotropic oceanic turbulence on the ERC is distinct. With the increase of propagation distance, these influence on the ERC gradually weaken, and can be ignored in the far field. In addition, z_R is always larger than that of the conventional Gaussian-Shell model beam in anisotropic oceanic turbulence. And these results will be useful in optical underwater communication.

The high-precision aspheric optics is difficult to achieve batch production because of its high complexity and cost in manufacturing process. Bonnet polishing is an efficient and precision polishing technology that has been widely used in aspheric optics manufacturing process. The existing aspheric bonnet polishing is mostly based on the five-axis CNC machine, but its cost is high. With the development of robot technology, it is a new, economical and effective attempt to combine robotic technology with bonnet polishing technology and apply it to manufacture the precision aspheric optics. In this paper, the research on robot-based bonnet polishing technology of aspheric optics is carried out. Firstly, the robotbased bonnet polishing equipment is introduced. Then, the control model of the robot-based bonnet polishing technology of aspheric optics is established. Finally, the aspheric polishing verification experiments are completed, and the good results are obtained. In this paper, the high-efficiency and high-precision manufacture of aspheric optics by robot-based bonnet polishing technology is realized, which greatly reduces the cost of aspheric precision manufacturing equipment, and provides a new technology choice for the high-precision, low-cost and batch production of aspheric optics.

SiO₂ aerogel has the characteristics of high porosity and high specific surface area. In order to explore the feasibility of using it as a new infrared functional additive for smoke composition, SiO₂ aerogel was mixed with NaNO₃ and NH₄NO₃ saturated solution respectively, and freeze-drying was introduced to make NaNO₃ and NH₄NO₃ deposit in the pores or surface of SiO₂ aerogel. So the powder-form infrared functional additive was prepared completely. The new powder not only reduces the frequency of the phenomenon of harden synthesized compound and improves its preparation process, but also guarantees the recovery of prefabricated particle size of SiO₂ aerogel in the combustion process. The experimental results show that: when mixed with smoke composition, on the condition that the mixing ratio of additives is 20% and negative oxygen balance is 20, compared with the traditional smoke composition obscuring ratio, the smoke composition synthesized obscuring ratio can be increased by 72% and 47% in infrared band.

With the development of modern industry, the monitoring and control of welding arc plasma is of importance than ever before. In order to capture the real-time image information of arc plasma in arc welding, the experimental system for high speed acquisition of arc plasma information is built in this paper. The experimental system is composed of high-speed camera, narrow-band filter and neural filter. The real-time image of metal argon arc welding of Q235 was captured by using high speed image system. The center wavelength of the narrow-band filter is 405 nm. The pseudo-color image and isothermal diagram of the arc plasma were obtained with software MATLAB. The radial temperature distribution of arc plasma was calculated with the Abel’s inversion. The results show that the temperature curve presents bell-shaped envelope. The maximum temperature occurs in the center of the plasma column and the value of temperature decreases gradually from the center. The experiment results lay foundation for the real-time monitoring and control of welding arc.

Linear mixing model is widely used in hyperspectral images unmixing for its simplicity while the real ground distribution does not satisfy the model. To achieve a high unmixing performance, the nonlinear spectral mixing model should be further discussed. This paper extends the spectral library by adding virtual endmembers based on the second scattering between endmembers to the original spectral library and transforms the nonlinear problem into a linear issue. By introducing the l₁ - l₂ sparse constraint and enhancing the virtual abundance weight, the optimization problem is solved by alternating direction method of multipliers. Experiments conducted with simulated data sets and real hyperspectral data show that the proposed algorithm is superior to other state-of-the-art methods.

In this note, we present our results about the process design and characterization of deep reactive ion beam etching of high line density fused silica transmission gratings. Up to 1 μm deep and 1880 lines/mm microstructures have been successfully fabricated for 532nm ultrashort pulse laser pulse compression using the developed processes. Considering the requirements of the manufacturing process, a certain range of parameters needs to be determined to reduce the need for etching precision. Therefore, in this paper, the strict coupling wave method is used to determine the parameter range of the grating in the range of the calculated grating parameters, and the diffraction efficiency of the simplified mode method can be calculated. To our knowledge, this is the first note on the design parameters of transmission pulse compression grating made on fused quartz substrate are given using the simplified mode method to design and a non-ICP-based reactive ion etcher.

With the highly accurate calibration requirements of ocean remote sensing, it is crucial to provide long-term in-situ measurement and validation for on-orbit remote sensors by using sea-based validation sites. However, when the laboratory-calibrated spectroradiometer is transmitted to in-situ validation sites, the measurement results are greatly deviated due to the difference between the field and the calibration environment, which directly affects the accuracy of on-orbit remote sensors synchronous calibration. In this paper, both of above water measurement and undersea profile measurement are considered. The temperature effects and stray light characteristics of common spectrometers were studied. In addition, the characteristics of the classic calculation model and optimization calculation model of immersion factor are compared and analyzed.

This paper proposes an improved neural network model based on CNN for evaluating the aesthetic quality of images. Firstly, we create three input sources from different angles of the target image and establish a three-channel parallel network model. Secondly, mlpconv layer is used to replace the traditional linear convolution layer to obtain more nonlinear abstract features in the network model based on CNN; Then, the combination of the global average pooling and full connection layers are used to replace the full connection layer in traditional CNN, and the three-channel features are merged. Finally, the EMD function is used as the loss function in the softmax layer. The output is probability density mass function from 1 to 10, and the mean and variance are used as objective qualitative score of picture quality. Experiments show that the proposed algorithm is feasible and effective, which solves the problem that the traditional method only obtains the binary classification of aesthetic. And this method gives the objective quantization score of the image. At the same time, the algorithm can get the evaluation value which is consistent with the actual situation in the real-time aerial experiment.

In this contribution, we fabricate a kind of part-doped yb fiber with 30/400μm core/cladding diameter by Modified Chemical Vapor Deposition (MCVD) in conjunction with chelate gas deposition technique. The yb ions doping diameter is 20μm in the fiber core and the core numerical aperture (NA) is 0.061. The fiber with a 50cm coiling diameter is characterized in a master oscillator power amplifier configuration, 2.42kW laser output with an excellent beam quality factor (M²~1.32) is experimentally obtained. For our best known, it is the first time to fabricate part-doped yb fibers by MCVD with chelate gas deposition method.

In order to solve the problem of matching failure of BBS (Best-Buddies Similarity) algorithm when the target image has a partial occlusion, cluttered background, imbalance illumination, and nonrigid deformation. A multi-feature template matching algorithm based on the BBS algorithm is proposed in this paper. On the basis of the location features and appearance features, we add HOG (Histogram of Oriented Gradients) features to make full use of the color, position and structural contour of the target image to match. In addition, we also perform mean filtering on the confidence map. The experimental results show that the AUC (Area Under Curve) score of the proposed algorithm is 0.6119, which is 6.38% higher than the BBS algorithm. Moreover, our algorithm has stronger robustness and higher matching accuracy.

Infrared point source target recognition is the target recognition technology based on non-imaging time-domain signal characteristics of infrared detectors for small size or long distance targets. It has wide application prospects in many fields, such as industry and national defense. However, infrared point source targets recognition has become a recognized problem, because this kind of targets has less information and more interference factors. Therefore, the research on infrared point source target recognition has important theoretical and practical value. This paper proposed a mathematical construction method of infrared point source target signal .And through the modeling of sensor response and the verification of hardware-in-the-loop simulation experiment, a set of time-domain signal data for infrared point source target recognition method is generated. Then, by extracting several time–domain features, a nearest neighbor classification method based on feature weighting is proposed by combining the idea of mathematical model matching. At the same time, the influence of the uncertainty of the model itself on the recognition effect is considered. The result shows, using the sample data of point source target generated by simulation, the best recognition time of proposed methods is approximately 6-8 seconds, and the recognition accuracy rate is 64.6%-79.2%.

Crop identification is important to precise agriculture, traditional survey is laboring intensive and classic remote sensing image processing methods are highly dependent on expertise, aiming at this problem, a U-Net based deep segmentation network is proposed for crop identification in UAV remote sensing image segmentation. First, we preprocess the UAV remote images of dataset for training. Then, we replace the transposed convolutional layer in U-Net upsampling block with sub pixel convolutional layer, to solve the problem that edge structure information lost and the low segmentation accuracy. The experiment result shows that, compared with original U-Net and other state of art segmentation model, the proposed method can improve the edge smoothness and pixel wise accuracy. The method uses sub pixel convolution layer for upsampling can be applied to other segmentation model, and the way uses segmentation model to identify the crop species can be extended to other remote sensing image processing filed.

The parallel wire fine line control system is mainly used for the adjustment of the distance between the two electrode wires and the parallel control during the fabrication of the miniature thermocouple. It mainly adopts precision electronic control and computer vision precision micro-measurement technology to ensure the adjustment and stability of the two-electrode wire spacing parameters during thermoforming. Hanging 17μm-60μm wire on the device, tightening the suspension wire by controlling the two-phase stepping motor, adjusting the five-phase stepping motor to adjust the parallel wire spacing, collecting the wire image in real time, automatically interpreting the wire spacing parameter by using a fast algorithm, and interpreting the wire The result output is displayed.

In recent years, along with the computer operation speed unending enhancement, the computer is employed to carry on the dangerous cargos the examination and the recognition to obtain the more and more widespread applications. Aiming at the disadvantage of high false detection rate in target classification detection using existing feature training classifiers, the work proposes a detection algorithm for hazardous articles with convolutional neural network on the basis of deep learning. For the image to be checked, sliding windows of different scales are used to determine whether there is an object window. For object detection, a convolutional neural network is trained with a large number of positive and negative samples. In order to better adapt to object detection, the topology of the convolutional neural network is improved. The window of suspected hazardous article is input into the improved convolutional neural network for dangerous object detection, and the false detection rate is reduced while maintaining the original detection rate.

Conical diffraction of a grating differs from the classical type of diffraction because the incident and diffracted wave vectors are not orthogonal to the direction of the grooves, and the light is almost parallel to the grooves. Compared with a grating in the classical type of diffraction, relatively higher diffraction efficiencies will be observed in conical diffraction. And, when the incident beam is perfectly parallel to the grooves of a rectangular grating profile (laminar grating), the symmetry of the setup causes diffraction of the intensity symmetrically around the plane of incidence. A multilayer grating is a grating which is coated with multilayers and can enhance the energy transport efficiency, especially for the tender X-ray range (1-8 keV) that covers a large number of K- and L-edges of medium-Z elements, and M-edges of high-Z elements. A multilayer laminar grating used in the symmetry of conical diffraction can obtain higher efficiency and has the feasibility of the amplitude beam splitting in the tender X-ray range. In this work, using numerical simulation, a possibility of multilayer laminar grating used in the symmetry of conical diffraction for beam splitting in 4.51 keV (Ti Kα1) has been demonstrated, showing a high efficiency and good flexibility of design.

Due to the difference of scenes, the uncertainty of detection for line features in stereo visual odometer reduces the accuracy of initialization for line features, and the point-line fusion visual odometry is easy to cause uneven distribution of features in feature extraction and matching process. The problems, which will lead to incomplete feature extraction and thus inaccuracy of pose estimation. In order to solve the problems, we design a stereo visual odometry system called PLWM-VO, which adaptively allocates the weights of the point features and the endpoints of line features in each frame through the adaptive weight assignment model. Firstly, the improved SAD algorithm is used to accurately detect the endpoints of line features, through which improves the registration accuracy of line features’ endpoints in the left and right, and improves the accuracy of line features in the initialization process. Secondly, the adaptive weight assignment model based on region partition assigns weights to point features and endpoints of line features. In this process, the effect of uneven feature distribution on the accuracy of pose estimation is reduced by region growth. Finally, experiment results on the open dataset shows that the proposed algorithm exhibits higher precision and stability than the state-of-the-art methods, which improves the positioning accuracy of our designed stereo visual odometry system.

Spectral imaging technology is an imaging technique based on narrowband spectroscopy developed in the 1980s. The spectral imager can simultaneously acquire the image information and spectral information of the target, and then accurately measure and analyze the composition of the substance. From food production to environmental monitoring, from material analysis to clinical diagnosis, from industrial monitoring to aerospace remote sensing, imaging spectrometers have become an important means for humans to obtain information, and there is a growing demand for civilian and military applications. It develops toward more spectral channels, higher integration, smaller size, and lighter weight. In this paper, the development methods and development status of several multi-channel spectral filters commonly used at home and abroad, as well as the research results of gradient filter design with half bandwidth of 10 nm in the 400nm~900nm band are introduced in detail.

Optical tomography imaging is widely used in target-detection, aerospace precision instrumentation and geological material detection for its non-contact, long-distance and high-precision imaging characteristics. Due to the different application range and structural design of the tomography system, the inevitable inadequate projection data and the offset of the rotation center may occur in the actual acquisition, which may cause artifacts and unclearness in the reconstructed image. In this paper, based on the research of image reconstruction algorithm, the paper compares filter back projection algorithm with iterative algorithm and analyzes effects of the reconstruction process with iterative algorithm under multiple parameters. Determining the appropriate weighting model, iteration number and relaxation factor, etc. Combined with high quality initial image and convex set constraints, an optimized SART algorithm is proposed. The experiment uses the optimized SART algorithm for image reconstruction. By comparing the image evaluation parameters with sharpness and average gradient, it is verified that the construction image with the optimized SART is better and clearer than those with the unoptimized SART and the simple filtered back projection algorithm.

Cognitive space-time adaptive (STAP) radar system optimizes the receiving filter according to the time-varying clutter environment. Aiming at the optimization design of receiving filter, an optimization model of STAP radar filter is established, and a method of filter optimization based on Lagrange dual algorithm is proposed. Compared with the traditional adaptive STAP filter, the simulation results show that the performance of cognitive reception STAP filter is analyzed and verified. Compared with the traditional adaptive STAP filter, cognitive STAP filter can significantly suppress clutter and effectively improve the signal-to-noise ratio.

Atmospheric aerosol refers to the stable mixing system composed of solid and liquid particles evenly dispersed in the atmosphere, and the particles are collectively referred to as aerosol particles. Aerosol particle detection is an important aspect of environmental monitoring. With the development of computer technology and holographic technology, the resolution of holographic measurement technology continues to improve. At present, digital holography can be used to measure the tiny displacement, deformation and vibration of opaque objects. In this paper, aluminum oxide powder is used to simulate aerosol particles. Aerosol particles are recorded by digital holography and holograms of aerosol particles are obtained. Finally, the shape and size of aerosol particles can be obtained.

The film laser concept has attracted much attention from both fundamental and practical points of view. However, to obtain a very thin film encounters the difficulty in realizing ultrahigh doping level of rare earth ions in a very thin film. In this work, a series of transparent Er³⁺/Yb³⁺ co-doped Al₂O₃-SiO₂ / P₂O₅-SiO₂ glass thin films with rare earth ions doping level up to 20 mol% were fabricated by sol-gel spin-coating method. The roughness value of film surface was characterized with AFM technique. The luminescent spectra and the corresponding FWHM of Er³⁺/Yb³⁺ co-doped films are compared and some other spectroscopic properties are discussed.

The main features of ship targets in infrared scene are low contrast, low signal-to-noise ratio and little detail information. In view of these characteristics, an infrared image enhancement algorithm based on gray interval segmentation is proposed in this paper. This algorithm expounds the imaging principle of infrared ship image from the perspective of radiology. According to the gray distribution of the image, the original image is segmented into gray interval sub images by using the gray histogram of the infrared image, and then the details of each sub image are enhanced separately. Finally, the final details of each sub image are enhanced by weighted linear superposition. The results show that the edge brightness and contrast of the image are obviously improved by the proposed algorithm. Compared with the traditional gray level compression method and the global histogram equalization method, the over-saturation phenomenon is obviously improved, and it is easy to realize.

During the capture operation, the hand-eye camera measures the spatial position of the center of the docking ring. However, the center can’t be used as a direct capture point, so we need to find a suitable capture point on the docking ring. This point can be connected with the center of the docking ring while satisfying the capture of the manipulator. This paper presents a method to find the capture point based on the intersection of the circle and a straight line, consisting of the image main point and the center of the circle. Furthermore, an algorithm for calculating the spatial position of the capture point is proposed. The digital and semi-physical simulation experiments verify the effectiveness of the method. The results show the position error of the capture point is within 0.8mm, and the capture angle error is within 0.4°.

With the continuous expansion of hyperspectral application scenarios, the traditional universal hyperspectral data processing software system is difficult to meet the needs of the industry, and cannot be quickly connected to the intelligent processing algorithm developed by the industry, which has become one of the bottlenecks in the promotion of hyperspectral to practical applications. In order to meet the needs of various industries for professional processing of hyperspectral data, fast access to intelligent processing algorithm, and highly efficient and reliable transplantation of intelligent processing algorithm to airborne platform, this paper designs an airborne hyperspectral data processing platform compatible with intelligent processing algorithms. The software architecture of "Platform + Plug-in" is realized, which provides comprehensive support for hyperspectral image processing and enables users to focus on the development of intelligent processing algorithms, which can be compatible with different intelligent processing algorithms through simple configuration.

X-ray pulsars were used for spacecraft navigation at the beginning of the discovery as their super stability spin periods (periodic changing rate 10^-19-10^-21). It is hopeful to get rid of the support of ground facility systems, which can be a real sense of spacecraft autonomous navigation in the large temporal and spatial range. Considering the pulsars radiation luminance, periodic instability and time of approach (TOA) uncertainty, the autonomous navigation accuracy can be achieved the order of kilometer magnitude. This paper reviews the X-ray pulsars navigation principle and the navigation missions in orbit in recent years. Finally, the study progresses of the X-ray pulsar telescope in our team are introduced. These efforts can be usefully for autonomous navigation in the future deep space exploration.

Aiming at the target characteristics and scene generation requirements of radar targets in defense system, a radar target attitude inversion method based on narrow-band radar tracking data is proposed. The measured data of the radar sensor is collected, the dynamic scene is analyzed and excavated, and the target radar line-of-sight angle inversion model is established based on the target electromagnetic scattering characteristic simulation data. Based on the model, the attitude inversion of the radar target is completed, and target dynamic RCS is used to verify the model. The correctness and effectiveness of the radar target attitude inversion method are verified.

Quantification of image clutter plays an important role in predicting target acquisition performances of a photoelectric imaging system due to the strong effect of optoelectronic image clutter. Accuracy in predicting the targeting performance of previous reported clutter metrics was relatively low because of disadvantages, such as lack of ability to accurately quantify the image with complex clutters and threshold selection problem. To address this problem, a novel multidirectionaldifference-Hash-based image clutter metric is proposed in this paper. Initially, an image similarity measure method based on multidirectional difference hash is established. Then, this method is applied to the quantification of image clutter, and an MDHash-based image clutter metric is obtained. Experimental results show that the proposed clutter metric correlates effectively with probability of detection, false alarm rate, and search time of observers.

Compare with the reflectance of land surface, ocean water is much less. And, the contribution of atmospheric molecules and aerosols plays an vital importance on the water-leaving reflectance inversion. In order to simplify the inversion process, we generate a look-up-table(LUT) that contains the observation geometry information, the aerosol optical depth(AOD), the exponent of Junge power law(V) and the other factors used to calculate the water-leaving reflectance. The AOD and V are determined using our previous iterative algorithm from dual near-infrared(NIR) and dual shortwave infrared( SWIR) channels, respectively. We compare the retrieved AOD and V with Aerosol Robotic Network(AERONET) measurement data to ensure the precision of aerosol information. The AERONET AOD at 550nm is 0.0876, and the inversed AOD from dual-NIR and dual-SWIR is 0.05-0.325 and 0.0373-0.98, respectively. For dual- NIR, there are 31.97% and 57.18% pixels with the AOD absolute relative error less than 10% and 20%, respectively. For dual-SWIR, there are 31.01% and 59.79%. Then, we use the retrieved aerosol information together with the observation geometry information to get the factors used to calculate the water-leaving reflectance through interpolation. Finally, we use the MODIS ocean color product to verify the water leaving reflectance calculated based on aerosol retrieved from NIR and SWIR, and the two calculated water-leaving reflectance are marked as ρNIR and ρSWIR. In the visible and near-infrared region, both of them are smaller than the product values. Despite the ρSWIR is larger than ρNIR, the overcorrection is much more serious in ρNIR.

Effects of oceanic channels degrade the performance of underwater optical communication (UOC) systems based on orbital angular momentum (OAM) multiplexing. Considering both oceanic turbulence and water attenuation, a more comprehensive channel model is proposed. We derive the expressions of bit error rate (BER) and aggregate capacity of UOC-OAM systems. We also investigate the system performance with wavelengths of 400-700 nm . The numerical simulation results show that the blue-green wavelength is not always the best choice to provide large channel capacities for UOC-OAM systems. For large chlorophyll-a concentration channel, due to the strong attenuation the optimal wavelength is shifted toward red wavelength region. Meanwhile, with increasing of oceanic turbulence strength, the frequency bandwidth decreases and systems with larger wavelength show better performance. These results will be beneficial to the design and application of UOC-OAM systems.

Reaction-sintered silicon carbide (RS-SiC) has been widely applied in space telescope mirrors, reflector, microelectronic mechanical systems due to its excellent properties. These applications require high surface accuracy and quality, but traditional mechanical contact polishing methods are hard to process this material due to the hardness and poor machinability. In this paper, atmospheric pressure plasma processing (APPP) which is a non-contact optical manufacturing technique with highly efficient and subsurface damage-free has been proposed to process this material, and optimal process parameter mainly about influence on removal function for machining processing RS-SiC have been found. The sample was observed and analyzed by scanning white-light interferometer (SWLI), which showed the calculated Peak- Maximum removal rate (MRR) and Volume-MRR were 12.526μm/min and 0.1298mm3 /min. Meanwhile, the surface chemical composition of RS-SiC was also investigated by XPS to reveal the plasma etching processes. The results illustrated that small number of radicals CxFy were introduced onto the RS-SiC surface during the plasma process which could be generated during the process of the reaction gas CF4 being excited. The process parameters and analysis of surface chemical composition in this paper will guide the further processing of RS-SiC.

Aiming at the need of multi-radar multi-view wide-band data fusion to improve imaging resolution, a data correlation and sparse component analysis method is proposed to analyze and coherently process multi-view multi-band broadband data in a small angle. The target characteristics of more viewing angles within the small angle range improve the target two-dimensional imaging quality. The simulation experiment analyzes the error of wideband data reconstruction and verifies the correctness and effectiveness of the data fusion method.

Proposed a dynamic decomposition technique for moving target observation task based on user requirements in the multi-satellite joint planning system. Carried out the research of moving target search algorithm based on dynamic update of grid target distribution probability. Realized the task decomposition of moving target observation supported by space-based information in the phase of monitoring and target searching. The key point of the algorithm is to consider the combination of multiple satellite observation bands in the process of multi-type satellite coordination, so that the coverage observation probability of multiple satellite observation bands is the maximum.

The focusing of hard X-ray is limited by the ability to design and to fabricate high-quality optics. The gains in resolution are often made at a sacrifice of focusing efficiency. The short Kinoform lenses offer a compromise position of high resolution and efficiency. Here we describe the design and the properties simulation of a series of short Kinoform lenses. After optimizing of the main characteristics, the focus spots size of the designed short Kinoform lens were proved to sub 500nm at photon energy of 30 keV.

We built an underwater photon-counting wireless optical communication (WOC) system, proposed and verified a video transmission protocol for high bit error rate (BER) environment. At the transmitting terminal, each pixel of the video frame has a sequence number indicating its position, and the sequence number information and the pixel information are encapsulated into independent data frames. After the receiving terminal distinguishes between the sequence number frame and the pixel frame according to the flag bit, the pixel and the sequence number are correspondingly placed into the storage unit. The experimental results show that when the BER is more than 0.6, the contour of the video frame can still be recovered, which effectively solves the problem of pixel dislocation of video frame caused by bit error.

Generative adversarial networks (GANs) has proven hugely successful, but suffer from train instability. The recently proposed Wasserstein GAN (WGAN) has largely overcome the problem, but can still fail to converge in some case or be to complex. It has been found that the use of weight clipping in WGAN to enforce a Lipschitz constraint on the critic, is the cause of the failure. We modify network architecture: use domain constraint layer instead of the use of weight clipping in WGAN. Experimental results show that our proposed method generates higher quality images than WGAN with weight clipping. And architecture is sample. Beside the network is more stable and easier to train.

In order to solve the inconvenience caused by the difference of hardware and software test resources during product testing, realize instrument interchange, meet the rapid change characteristics of test demand and rapid deployment requirements, this paper studies the construction method of a distributed test environment, adopts a fixed general test platform in hardware, and drives Heterogeneous hardware interface standardization technology (middleware technology), isolation module differences, to solve the different manufacturers to the module replacement to bring compatibility problems. Software adopts the design principle of layered software, connecting test elements such as device libraries, ICD libraries, use case libraries, etc. through the use case design layer, resource allocation layer, and hardware environment layer, and through XML between elements Exchange data information, interact with database information via Ethernet, and complete the automatic testing process. Each software runs independently on multiple test devices in the test environment, enabling the automatic execution of corresponding actions or automatically generating the test case software code through XML file.

By virtue of the characteristics of high machining certainty, fast surface shape convergence and less subsurface damage, magnetorheological finishing (MRF) has been widely applied in ultra-precision machining for optics. However, the convolution of tool influence function (TIF) and the regular movement track make it difficult to avoid the deterioration of mid-spatial frequency (MSF) errors in the MRF process, which will affect the optical properties of optics. In this paper, the main factors affecting the MSF errors in grating path processing are theoretically analyzed. Under the condition of determinate polishing spot, the main influencing factors are the removal depth and processing spacing of a single scan. Through experimental research, the influence of removal depth and processing spacing of a single scan on the MSF errors in the MRF process is acquired. On this basis, the preferred parameters of removal depth and processing spacing of a single scan under specific processing conditions in the MRF process are obtained with the MSF errors of 1 nm as the evaluation index. It provides a theoretical basis and reference value for suppressing the deterioration of MSF errors of optics during MRF process and obtaining high quality optical surfaces.

Safety test has an important impact on the suspension management system, this paper starts from the top-level requirements of the product under test, according to the functional performance characteristics of the suspension management system, to identify possible factors affecting safety, design-related tests to verify, so as to improve product safety. Based on the definition of weapon security of suspension management system, based on the existing demand analysis model and theoretical methods to extract the top requirements of suspension management system and product safety design requirements, combined with the actual project testing experience, build a suspension management system weapon security demand analysis method, Combined with specific test project analysis of the actual use scenario of the product under test, testers can use this analysis template to quickly identify the key safety-related events in the product under test, form specific test requirements and design test cases.

Segmentation of anatomical structures in ultrasound images required radiological technology and a great deal of ultrasonic experience. The manual segmentation is often dependent on expertise of clinicians and time-consuming. Therefore, we present an automatic system for segmentation and measurement of ultrasound images. We propose a scale attention feature pyramid network (SAFNet) for fetal biometric measurements from two-dimensional ultrasound images. The scale attention module is steered to form feature pyramid at each level. Auxiliary layer is used to learn object boundary definition with deep supervision. Further, we present a two-stage framework which is an automatic classification measurement system (ACMS), firstly classifies the image type which has three labels: head, abdomen and femur. Then outputs the final segmentation result. The SAFNet results better performance on our datasets compared to the baseline U-Net. Experiments show that the ACMS results in classification accuracy of 95.27%/90.94%/94.93% of fetal head, abdomen and femur test set, respectively. Feature pyramid and attention mechanism inside the network for feature selection results in improvement in the segmentation accuracy. The ACMS can conveniently obtain segmentation result no matter what type is given.

The atmosphere is full of liquid and solid particles, both man-made and natural. Atmospheric particles play a key role in weather, climate and biological health. Therefore, the accurate measurement of atmospheric particles is of great significance. Digital holography, as a new generation of imaging technology, has the advantages of simple and stable structure, three-dimensional imaging, and the ability to distinguish laminated particles. Digital holography technology has been widely used in the field of three-dimensional morphology measurement, interferometry, nondestructive testing, holographic microscopy, particle field measurement and other fields. In the field of particle field measurement, most optical particle field measurement technologies can only measure spherical particles to obtain the statistical distribution of particle concentration and diameter, such as laser scattering method and phase doppler method.Digital holography can obtain the morphology details, particle size parameters and three-dimensional spatial distribution spectrum of each particle in the particle field. On the basis of the original device, the two-views hologram is recorded by adding a reflector. The coordinates of the particles along the optical axis will be determined by the position of the reflector and the position of the reflected image.

Aiming at small objects detection such as unmanned aerial vehicle (UAV), this paper proposes a fast object detection algorithm based on depth wise separable convolutions. Firstly, the inverted residuals units based on depth wise convolutions and pointwise convolutions are used to construct a lightweight feature extraction network to improve the network’s speed. Secondly, the feature pyramid network is used to detect the five scale feature maps to improve the detection performance of small objects. Otherwise, we make an UAV dataset based on the urban background for training and testing of our experiments. The experimental results show that the improved method proposed in this paper can effectively improve the detection accuracy and real-time performance of UAVs in complex urban backgrounds, and the computation of network is greatly reduced, thereby making it possible to achieve object detection on embedded systems.

Convolution Neural Networks (CNN) and Recurrent Neural Networks (RNN), which are the main research methods of image caption, have developed rapidly. Nevertheless lacking of global consciousness in image caption has not been completely solved. Separation from the bottom-up visual attention mechanism and the top-down visual attention mechanism has been widely used in image description and visual question and answers. In this article, we put forward image description based on a global attention mechanism research methods. The global attention prior channel is added to the infrastructure to extract the global information features while learning the local features. The attention to the object and other outstanding level of image region is calculated, and the global image features are enhanced.

According to the current situation of fire detection and the requirements of early fire detection, a new method is tried to detect fire in order to realize more accurate fire detection.That is to apply hot microholography to fire detection.A fire detection experimental system based on mathematical microholography is designed in this paper.The main theoretical basis of microscopic holographic technology can detect the three-dimensional model of fire smoke particles, and then determine whether the fire occurs or not by pattern recognition algorithm.

LCD TVs are developing in the direction of large size and ultra-thinness. The glass substrate is subject to the influence of gravity caused by the increase in size and the problem of backlight heating, which in turn affects the optical characteristics of large-size LCD TVs and causes display defects. Therefore,this paper explains the large size from the perspective of the change of light transmittance of liquid crystal panel by using the liquid crystal panel mechanical test and establishing 65-inch LCD TVs model and optical simulation model, using the combination of finite element simulation and TechWiz LCD optical simulation. The LCD TVs displays the defect mechanism. Finally, when the tilt angle is less than 0.5 ° and the operating temperature is lower than 40℃,the liquid crystal panel can effectively reduce the display defects of large-size LCD TVs.

Traditional hyperspectral feature extraction methods focus on spectral features and neglect spatial features,its extraction method is set in advance and is not suitable for all hyperspectral images. Faced with these problems, we propose a three-dimensional convolutional network for hyperspectral classification, which consists of a convolutional layer,2 downsampling layers, 2 identification layers, a flatten layer, and 4 fully connected layers. The proposed network employs three-dimensional convolution operation to extract spectral-spatial features from hyperspectral images,there are two reasons for this, the first reason is three-dimensional convolution can automatically learn a large number of mappings between input and output.The second reason is three-dimensional convolution can effectively extract spectral-spatial features and improve network classification performance. In order to extract high-level features and prevent network performance degradation, the proposed network adopts residual connections.More importantly, the OpenMax algorithm is employed to detect hyperspectral unknown targets. In addition to the probability that the output belongs to a known class, the OpenmMx adds the probability that the predicted input belongs to unknown classes, as a result,the deep convolutional network can respond to inputs of unknown classes.experiments based on typical hyperspectral data show that the proposed network perform accurately in the known classes classification and the openmax algorithm is suitable for unknown targets detection of hyperspectral images.

In order to understand the stimulated Brillouin scattering (SBS) process in water, it is necessary to analysis theoretically the factors which to determine the SBS gain coefficient and threshold. The coupled wave equation is used to investigate the effects of temperature, salinity and attenuation coefficient on SBS gain coefficient and threshold. The results indicate that the higher the temperature and salinity, the greater the SBS gain. And the threshold is negatively correlated with the temperature. The effects of attenuation coefficient on threshold of steady-state SBS is greater than transient SBS.

BEEPS(bi-exponential edge-preserving filter) is used to enhance the details of infrared image in this paper. The original infrared image has a dynamic range of 12 or 14 bits, and the human observation range is only 8 bits. Usually, the original infrared image needs to be compressed and displayed by gray-scale remapped for displaying. For example, automatic gain control and histogram equalization are the most widely used image display technologies in infrared imaging systems, but they can lead to the loss of local details, and it is difficult to control the visibility of weak details in images. Therefore, an infrared image digital detail enhancement algorithm has emerged. Current digital enhancement algorithms can effectively enhance image details and avoid over-amplification of noise, but there are still some drawbacks, such as large computational load and poor application flexibility. Therefore, we use BEEPS in our algorithm to overcome these problems. This algorithm uses a two dimensional convolution to separate the detail information from an original infrared image, and turn the original image into the detail layer and the base layer. Detail layer processing is to transform two-dimensional convolution into one-dimensional convolution, and to complete one-dimensional convolution through iterative calculation. Then, the enhanced detail layer is added back to the base frequency layer of histogram equalization. This not only improves the computational efficiency, but also improves the visual quality of the original image. The BEEPS algorithm is proved to be excellent by image and data testing.

Aiming at the characteristics that the contrast of schlieren sphere image is weak, the color of sphere target is black, and the sphere target is surrounded by background area in the fast collimation process of integrated diagnostic system, a method for calculating the schlieren sphere center based on gray stretching and digital morphology is proposed in this paper. Firstly, the original image is transformed by linear gray scale to enhance the contrast between the schlieren sphere target and the background area, and then the linear gray scale image is binarized by using the OTSU method. Secondly, the background area is transformed into a continuous connected region by using digital morphological operations such as corrosion and expansion, and search for the maximal connected area is the background area. Thirdly, the image of using digital morphology method and the image of background area are processed NOR operation to obtain the schlieren sphere target. Finally, on the basis of edge detection by sobel operation, the least square method is used to fit the center and radius of the schlieren sphere. The experimental results show that the algorithm can calculate the center of the schlieren sphere whose gray level difference between the sphere area and the background area is greater than 10( the max gray level is 4095). The error between the center coordinate obtained by circle fitting method and the real center coordinate is less than 2 pixels, which meets the requirement of the integrated diagnostic beam fast automatic collimation system that the accuracy of the schlieren sphere center is less than 3 pixels.

The research of artificial intelligence use the basic theory, method and technology of computer software and hardware to simulate human intelligence behavior. With the development of artificial intelligence technology, artificial intelligence plays an important role in more fields. Artificial intelligence emancipates the productive forces, while simplifies manual operation steps and improves work efficiency. Medical imaging measurement is a professional work, and traditional methods are restricted by operators, equipment and many other aspects. The introduction of artificial intelligence provides a new way for the development in this field. In this paper, the actual needs of fetal ultrasound image measurement are analyzed. the image intelligent measurement algorithms of different parts are introduced in detail. their advantages and disadvantages are analyzed, and the development direction of fetal ultrasound image measurement is prospected.

Water depth is an important marine elements, and one of the main contents of marine surveying and mapping. Compared with traditional single-beam and multi-beam sonar measurement methods, remote sensing method has become an important supplementary method for bathymetric data of the area that the vessel can not be directly reached，due to its advantages of large-scale and low-cost.
Sun glitter is an interference factor of shallow water depth multispectral remote sensing, and it contains water depth information. Therefore, how to separate the sun glitter information and water radiation information, and improve the inversion ability of water depth using the two parts of information, is a scientific and practical research topic.
In this study, the Lingshui Bay of Hainan Island was selected as the main study area, and the sun glitter separation method of high resolution multispectral remote sensing images and the semi-empirical water depth inversion models are evaluated and analyzed. And a multi-spectral remote sensing inversion method for water depth without the support of measured water depth data was developed. The following research results are obtained:
The semi-empirical water depth inversion models were used for inversion and evaluation analysis of the water radiation information obtained by different sun glitter separation methods. Results showed that the logarithmic ratio depth model with blue and green bands is the best in the shallow area of 10 m. When the solar radiation of remote sensing image is weak, the best water depth inversion results can be obtained using Martin algorithm, the R2 is 0.94, the RMSE is 1.27m. When the solar radiation of remote sensing image is strong, the best water depth inversion results can be obtained using Hedley algorithm, the R2 is 0.89, the RMSE is 0.94m.
A multispectral remote sensing water depth inversion algorithm without the support of measured water depth data was developed based on sun glitter information and water radiation information. The evaluation results showed that RMSE is 0.92m in shallow waters with depth less than 6m. Therefore, this algorithm has a certain application potential in shallow sea area without measured data.