This PDF file contains the front matter associated with SPIE Proceedings Volume 8334, including the Title Page, Copyright information, Table of Contents, and Introduction, and Conference Committee listing.

An regularization approach is introduced into the online identification of inverse model for predistortion. It is based on a modified backpropagation Levenberg-Marquardt algorithm with sliding window. Adaptive predistorter with feedback was identified respectively based on direct learning and indirect learning architectures. Length of the sliding window was discussed. Compared with the Recursive Prediction Error Method (RPEM) algorithm and Nonlinear Filtered Least-Mean-Square (NFxLMS) algorithm, the algorithm is tested by identification of infinite impulse response Wiener predistorter. It is found that the proposed algorithm is much more efficient than either of the other techniques. The values of the parameters are also smaller than those extracted by the ordinary least-squares algorithm since the proposed algorithm constrains the L2-norm of the parameters.

With the characteristics of high speed, heavy density and strong dependence on control equipment, safe high-speed railway operation are affected by various factors whose interaction relationships resulted in the complexity and high risks of high-speed railway. To adopt 'basic theory, key technology and simulation experiment' method, based on the "proactive safety" concept, a visual, intelligent and integrated monitoring and early-warning platform for safe high-speed railway operation was proposed in this paper, and was described from objectives, architecture and functions, to achieve comprehensive monitoring, safety warning, decision support and information transfer and sharing in the whole operation.

For the purpose of improving the defect and deficiency in the field of using software to process images with wavelet transform, this paper introduced one core algorithm in the newest static image compression standard JPEG2000, that is the reversible 5/3 integer wavelet transform. Then improved the algorithm and programming it based on nonstandard decomposition method with Visual C++. Eventually a multi-document platform which has a strong portability was created. 5/3 integer wavelet transform and inverse transform for grayscale can be carried out by this platform. The wavelet coefficients could also be stored, which can be used for deeper research, such as filtering, image encoding and decoding, digital watermark and so on. What's more, after processing, the wavelet coefficients might be displayed as an image, which makes the results more intuitive. In addition, we illustrated the application of the platform through an example. After verification, 5/3 integer wavelet transform for images is indeed reversible and lossless, and the original image could be accurately reconstructed. The platform can be also applied to other image processing fields.

In this paper, a new concept that is Circulation Course Resources (CCR) is introduced, which means the course resources circulating from the students listening in classroom lecture, camera shooting, video coding, video storage, video server to the students learning from VOD. The creating course video system and network-teaching system as parts of CCR architecture are presented separately. To connect the two systems, a middle system defined as Bridge System is designed and modeled with UML. The core application design of the Bridge System is expressed by the classes design and main database design. The functions of Bridge System include making the course videos flowing from one system to another automatically and converting the important data of the two systems into uniform format. The CCR architecture has been put into action and achieved satisfied results.

PCA, LDA and LPP are the three most representative subspace face recognition approaches. In this paper, we show that they can be unified under the same framework. A unified framework is constructed by using the graphic embedding. We develop a unified framework to study the three major subspace face recognition methods: PCA, LDA, and LPP. PCA is an evaluation benchmark for face recognition. Both LDA and LPP have achieved superior performance in the YALE face database. A unified framework on the three methods will greatly help to understand the family of subspace methods for further improvement.

This study uses 3D virtual reality technology to create the "Mackay campus of the environmental education and digital cultural 3D navigation system" for local historical sites in the Tamsui (Hoba) area, in hopes of providing tourism information and navigation through historical sites using a 3D navigation system. We used Auto CAD, Sketch Up, and SpaceEyes 3D software to construct the virtual reality scenes and create the school's historical sites, such as the House of Reverends, the House of Maidens, the Residence of Mackay, and the Education Hall. We used this technology to complete the environmental education and digital cultural Mackay campus . The platform we established can indeed achieve the desired function of providing tourism information and historical site navigation. The interactive multimedia style and the presentation of the information will allow users to obtain a direct information response. In addition to showing the external appearances of buildings, the navigation platform can also allow users to enter the buildings to view lifelike scenes and textual information related to the historical sites. The historical sites are designed according to their actual size, which gives users a more realistic feel. In terms of the navigation route, the navigation system does not force users along a fixed route, but instead allows users to freely control the route they would like to take to view the historical sites on the platform.

This paper describes an image segmentation algorithm based on Non-subsampled Contourlet (NSCT) and Normalized Cut (N-Cut). NSCT have the characteristics of multiresolutional, localized, multidirectional, and low redundant and anisotropic, so it is more effectively capture high dimensional singularity in an image. The segmentation is accomplishment by using normalize-cut rules after NSCT. The experiment results show the efficiency of image contour extraction, and avoid the problem of over-segmentation and under-segmentation phenomenon in N-Cut algorithm. In addition, because of less information in the low resolutions, the processing time is largely shortened, and then gets better segmentation results.

An improved video median noise reduction algorithm is presented for 120 vehicle terminal monitoring system in this paper. The noise causes of the video images in many 120 ambulance vehicle video terminal monitor equipments are analyzed, and the space rigid body model of self-adaptive median noise reduction filter is established to decrease the noises of the video image transmission process. The noise reduction experiment of video images shows that the proposed video median noise reduction algorithm is superior to the traditional adaptive filtering method, because the new method has the superiority of space-time joint noise reduction.

A method which can distinguish tower cranes from other objects in an image is proposed in this paper. It synthesizes the advantages of both morphological theory and geometrical characters to identify tower cranes accurately. The algorithm uses morphological theory to remove noise and segment images. Moreover, geometrical characters are adopted to extract tower cranes with thresholds. To test the algorithm's practical applicability, we apply it to another image to check the result. The experiments show that the approach can locate the position of tower cranes precisely and calculate the number of cranes at 100% accuracy rate. It can be applied to identifying tower cranes in small regions.

This paper presents a fractal dimension calculation technique of highway pavement which is used to detect the pavement cracking. Firstly, the impulse noises generated by pavement unevenness are eliminated by the use of median filtering of pavement image. Secondly, the fractal dimension of fissure is calculated according to the crack's fractal feature. Finally, the pavement damage is detected and located with respect to fractal dimension. These three steps can realize the location detection and damage rate computation of pavement damage successfully. To a number of random images, the experiment results show that the fractal dimension of pavement crack region is from 2.25 to 2.99 basically.

This paper presents image detection method of pavement crack based on fractal dimension feature and designs self-adapting algorithm of fractal dimension interval of pavement region. Through image pretreatment, calculation of fractal dimension, self-adapting calculation of dimension interval, we obtain the location image of damage pavement. The experimental results of transverse crack, longitudinal crack, net-shaped crack, pit slot are contrast with that of Sobel operator. The results show that they have the similar capability on the representation of crack, but the proposed method is more flexible on the aspect of representation of crack size and calculation of damage ratio.

Currently, mobile image applications spend a lot of computing process to display images. A true color raw image contains billions of colors and it consumes high computational power in most mobile image applications. At the same time, mobile devices are only expected to be equipped with lower computing process and minimum storage space. Image dithering is a popular technique to reduce the numbers of bit per pixel at the expense of lower quality image displays. This paper proposes a novel approach on image dithering using 2x2 Tchebichef moment transform (TMT). TMT integrates a simple mathematical framework technique using matrices. TMT coefficients consist of real rational numbers. An image dithering based on TMT has the potential to provide better efficiency and simplicity. The preliminary experiment shows a promising result in term of error reconstructions and image visual textures.

With the development of computer education software, digital educational game has become an important part in our life, entertainment and education. Therefore how to make full use of digital game's teaching functions and educate through entertainment has become the focus of current research. The thesis make a connection between educational game and collaborative learning, the current popular teaching model, and concludes digital game-based collaborative learning model combined with teaching practice.

A HTTP based video transmission system has been built upon the p2p(peer to peer) network structure utilizing the Java technologies. This makes the video monitoring available to any host which has been connected to the World Wide Web in any method, including those hosts behind firewalls or in isolated sub-networking. In order to achieve this, a video source peer has been developed, together with the client video playback peer. The video source peer can respond to the video stream request in HTTP protocol. HTTP based pipe communication model is developed to speeding the transmission of video stream data, which has been encoded into fragments using the JPEG codec. To make the system feasible in conveying video streams between arbitrary peers on the web, a HTTP protocol based relay peer is implemented as well. This video monitoring system has been applied in a tele-robotic system as a visual feedback to the operator.

Noise in an image is the unwanted information present; it should be removed without disturbing the useful information present in it. De-noising an image is very active research area in image processing. In this study, three classes of degraded noise images are used they are additive noise, multiplicative noise and impulsive noise. There are several algorithms for de-noise but each algorithm has its own assumptions, advantages and limitations. Histogram multithresholding give rise to explicit peaks, which reduces the task for finding thresholds in dissecting the image histogram. The proposed method uses histogram multithreshold segmentation as the first step followed by statistical features and pattern classifiers for identifying the noise type. Simple filters are used to get the noise samples and noise identification is achieved by using the proposed method. The proposed method yields the higher results when compared with the first method for classifying the noise types.

Several algorithms have been implemented to resolve the problem of text categorization. Most of the work in this area geared for English text, whereas few researches have been conducted on Arabic text. However, the nature of Arabic text is different than English text; pre-processing of Arabic text are more challenging. In this paper an experimental study was conducted on three techniques for Arabic text classification; these techniques, Discriminative Multinominal Naive Bayes (DMNB), Naïve Bayesian (NB) and IBK Algorithms, The paper aimed to assess the accuracy for each classifier and to determine which classifier is more accurate for Arabic text classification based on stop words elimination. The accuracy for each classifier is measured by Percentage split method (holdout), and K-fold cross validation methods, along with the time needed to classify Arabic text.

An double Nd:YAG regenerative amplification picosecond pulse laser is demonstrated under the semiconductor saturable absorption mirror(SESAM) mode-locking technology and regenerative amplification technology, using BBO crystal as PC electro-optic crystal. The laser obtained is 20.71ps pulse width at 10 KHz repetition rate, and the energy power is up to 4W which is much larger than the system without pre-amplification. This result will lay a foundation for the following amplification.

We report on a picosecond pulse laser produced by grating stretcher and regenerative amplifier. By designing the experimental setup and numerical simulation, mode-locking output pulse energy of 7.5mJ with a repetition rate of 1KHz and a pulse width of 106.4ps stretched from 8.5ps is obtained at 1064nm. The results indicated that this system has laid a good foundation for the multichannel amplification to get higher pulse energy.

A micro processing used LD end-pumped Nd:YVO₄ all solid-state picosecond pulse laser was demonstrated under the semiconductor saturable absorption mirror(SESAM) mode-locking technology and regeneration amplifier technology, by using BBO crystal as electro-optic crystal and diode-side-pumped Nd:YAG. 1064nm laser was obtained with 1.47mJ single pulse energy, 15ps pulse width at 1 kHz repetition rate and the pulse energy fluctuation was less than 0.6% in 3 hours operation. Finally, through the galvanometric we got the beam focused, realizing the steel plate processing which thickness was 0.5mm and the aperture radius was 25.5μm.

Statistical approaches become very important tools that interfere and overlap in our daily life and become inevitable event that help us in every tiny details of our life, in this paper; we are going to present a new technique for analyzing the two principal component of any given object by calculating the direction over the occupied coordinates using mean, variance, and covariance statistical functions, and by finding some relationship between those statistical functions; we have extracted the angle degree of the processed object, for pattern recognition applications; this object can be adjusted accordingly to overcome the rotation perturbation shortcoming that hinders the extraction of a unified features especially for object recognition purposes in which we have to present many samples per single pose which makes the processing of this increasing size of the database is a noticeable burden, we have achieved a dramatically results with almost zero time of calculation since the statistical functions applied need little processing time to finish.

Remote sensing is a very useful method for data collection in open spaces, especially in precision agriculture and has been widely used over centennial. This paper presents the development of methodologies and identification of a surface model grasslands and pastures based on of chosen guidelines and properties. The model will be used to automate the process of monitoring the grasslands based on the analysis of spatial data and computer analysis of aerial photographs obtained automatically.

In this work a mesh-growing surface reconstruction method adapted for noisy and sparse data is presented. The method realizes an interpolating growing model and introduces the Complex Propagation scheme that integrates both inertial and tangential propagation terms. The features and basic functionality of the method are described. A number of experiments are carried out on 3D ultrasound data received from 2D tracked free-hand ultrasound and sequential triggered scanning systems.

This paper introduces a novel global thresholding approach that exploits the product of gradient magnitude (PGM). The PGM of an image is obtained by multiplying the responses of the first derivative of Gaussian (FDoG) filter at three adjacent space scales. The output threshold is selected as the one that maximizes a new objective function of the gray level variable t . The objective function is defined as the ratio of the mean PGM values of the boundary and nonboundary regions in the binary image obtained by thresholding with variable t . Through analysis of 35 real images from different application areas, our results show that the proposed method can perform bilevel thresholding on the images with different histogram patterns, including unimodal, bimodal, multimodal, or comb-like shape. Its segmentation quality is superior to five popular thresholding algorithms.

This paper present an improved imaging algorithm for bridge crack detection, through optimizing the eight-direction Sobel edge detection operator, making the positioning of edge points more accurate than without the optimization, and effectively reducing the false edges information, so as to facilitate follow-up treatment. In calculating the crack geometry characteristics, we use the method of extracting skeleton on single crack length. In order to calculate crack area, we construct the template of area by making logical bitwise AND operation of the crack image. After experiment, the results show errors of the crack detection method and actual manual measurement are within an acceptable range, meet the needs of engineering applications. This algorithm is high-speed and effective for automated crack measurement, it can provide more valid data for proper planning and appropriate performance of the maintenance and rehabilitation processes of bridge.

Stereo vision system is a useful method for deepness or depth gathering of objects and features in an environment. This paper presents the region of interest ROI in disparity mapping to be analyzed to get the estimate distance on stereo vision applications. The stereo vision application in this paper is using a mobile robot that navigates using a pair of camera. The cameras work as a stereo vision sensor for its navigation. The ROI is a reference sight of the stereo camera which the pixel intensities from the disparity mapping determine the distance or depth using an algorithm. The stereo vision baseline is based on horizontal configuration. The matching process is using block matching technique which briefly described with the performance of its output. The disparity mapping is generated by the algorithm with the reference to the left image coordinate. The algorithm uses Sum of Absolute Differences (SAD) which is developed using Matlab software.

To avoid the tremendous complication and difficulty of the digital image analysis method in the laboratory which is an important approach to study the microstructure of granular materials that plays significant role and govern the macroscale of granular soils, a new method to perform the digital image analysis of microstructure of the granular soils based on numerical simulation is proposed. A series of numerical models are developed to simulate the plane strain tests of granular soil. Based on the numerical results, two methods are proposed to simulate digital image method in the laboratory: the RENCI 3D slicer method which is a direct analogous to the laboratory method, and the geometric algorithm method which takes advantage of the numerical method and is considered to be more accurate and be able to get more information of the microstructure. Some analyses of the local void ratio distribution and the particle orientation distribution are performed as on physical laboratory experiment. The proposed numerical digital image analysis method is proved to be a valid and more efficient approach for stereological analysis of microstructure of granular soils.

Time-interleaved Analog-to-Digital Converter (TIADC) is an efficient way to achieve higher sampling rates for medium-to-high resolution applications. However, the performance of a TIADC suffers from mismatch errors among the sub-channels. This paper presents a method to estimate the channel mismatch errors using the sub-channels' output data. The proposed method introduces an input dependent estimating model (IDEM) based on an equivalent transfer function including the mismatch errors to calculate the standard deviation of the mismatch errors. The spurious-free dynamic range (SFDR) is then evaluated by applying multi-tone sinusoids signal to input. The simulated results show that the method in this work can get about 45dB SFDR enhancement.

According to the characteristic of low visibility of night image, we proposed an improved image enhancement method based on multi-scale Retinex. In our method, we replace the Gaussian filter as Blur filter, and replace logarithm gain as linear gain to reduce the computational complexity. The proposed algorithm first segment color image into three independent R, G, B channels, and then employ our improved multi-scale Retinex algorithm to enhance contrast of each channel. Finally, the enhanced image is given by recombining the three R, G, B enhanced results. A large number of experiments show that the proposed algorithm can quickly and efficiently improve the contrast of the night image and has a better visual effect than that of the similar method.

When a document is fed to a scanner either mechanically or by a human operator for digitization, it suffers from some degrees of skew or tilt. Skew detection is one of the first operations to be applied to scanned documents when converting data to a digital format. Its aim is to align an image before processing because text segmentation and recognition methods require properly aligned lines. This paper presents a new approach for skew detection and correction of documents. Particle Swarm Optimization (PSO) is used for solving skew optimization. A new objective function based on local maxima and minima of projection profile is formulated and PSO is used to find the best angle that maximizes the objective function. The proposed method is compared with existing methods like Hough transform, Fourier transform. Experimental results show that the proposed method corrects the skew with the maximum error rate of ± 1° .

The automatic target recognition based on image fusion refers to the fusion process using the target images provided by a variety of sensors, so as to improve the recognition accuracy and robustness and to obtain better recognition performance. This paper presents a feature fusion method for global structure information and local structure information of both the feature extraction and target classification recognition, the global structure information is obtained through the scatter matrix, while the local structure information is obtained through constructing the Laplacian matrix of the nearest neighbor graph, in the criterion function, the importance degree of these two information in different applications is regulated by setting a regulatory factor.

Landslide is one of prominent geohazards that continuously affecting the tropical countries including Malaysia. Frequent occurrences of landslides at hillslopes during the heavy rainy periods have resulted in public fear for the safety of their life and properties. For the past 25 years, many landslides have occurrences have been reported in Klang Valley especially at the hilly terrain residential areas. A landslide monitoring scheme is therefore very crucial and should be carried out continuously. Various studies have been conducted to monitor landslide activities such as conventional geotechnical and geodetic techniques. Each of these techniques has its own advantages and limitations. Therefore, this study focuses on the effectiveness of the combination approach of GPS technology and inclinometer techniques for landslide monitoring. The study area is located at residential area Section 5, Wangsa Maju, Kuala Lumpur, Malaysia. The inclinometer instrument has been placed at five (5) selected monitoring points and three (3) epochs of inclinometer measurements were made. At the same time, the GPS observations have also been carried out for three (3) epochs separately using GPS static techniques. This GPS network consists of four (4) control points and eleven (11) monitoring points. The GPS observations data were validated, processed and adjusted using two (2) adjustment software namely Trimble Geomatic Office (TGO) version 1.6, and GPS Adjustment and Deformation Analysis (GADA). The results have shown that the GPS technique can be implemented with inclinometer technique to detect horizontal displacements up to ± 30 mm and vertical displacements less than ± 50 mm.

This paper introduces a image retrieval algorithm based on SIFT feature. The image was transformed feature vecter. The similarity was computed using Euclidean distance about two images' feature. Experiments show that it has good performance on special object.

This paper presents a multilayer segmentation of stereo images with reference to the displacement in the left and right images. A stereo image is given as left and right components, the corresponding matching is found by drawing random parallel lines in x-axis making the y-axis as constant. The edges is found in both the right and left images with its pixel position. The number of edges found in two components is noted down. Then the edge values are clustered with respect to the deviation found in the matching correspondence. The rough distance is calculated using the deviation clusters. The number of clusters represents the number of layers of the segmentation. Once the layers are determined the whole image is segmented with zero crossing by taking the displacement as the layer parameter. The algorithm is implemented and tested for single and multiple objects with various distances in feet.

This paper presents a new and efficient method for the detection of concentric ellipses with the same orientation in images on the basis of the Hough Transform (HT). In order to meet real application requirements of high detection accuracy and low time and space costs, the method detects the parameters of ellipses separately with a five-step algorithm by using the special geometry properties of concentric ellipses instead of gradient information or central symmetric points. Some experimental results on both synthetic and real images show that the method is more efficient, accurate, and robust than the Randomized Hough Transform (RHT) method.

Driver fatigue is an important reason for traffic accidents. To ameliorate traffic safety, this paper proposes a novel method for fatigue pattern detection which is based on parallel Gabor and 1-Nearest Neighbor (1-NN) algorithm. In the algorithm, parallel Gabor wavelets ssand feature orientation fusion are first employed to get multi-scale orientation fusion image features, since facial features from tired drivers are different from the opposite. Then, during the phase of classification, the multi-scale 1-NN algorithm is used to classify the extracted facial image features for fatigue pattern detection. Experimental results show that the new method can effectively recognize driver fatigue pattern, and the performance of real-time fatigue detection with multiple processors has been improved comparative to single CPU computing environments.

Iris location plays an important role in iris recognition system. Traditional iris location methods based on canny operator and integro-differential operator are affected by reflections, illumination inconsistency and eyelash. In this paper, we introduce an accurate iris location method for low quality iris images. First, a reflection removal method is used to interpolate the specular reflection. Then, we utilize Probable boundary (Pb) edge detection operator to detect papillary boundary with a lower interference point. Moreover, we optimize the Hough transform to obtain high accuracy result. Experimental results demonstrate that the location results of the proposed method are more accurate than other methods.

In this paper, we propose a new secret image sharing scheme based on chaotic system and Shamir's method. The new scheme protects the shadow images with confidentiality and loss-tolerance simultaneously. In the new scheme, we generate the key sequence based on chaotic system and then encrypt the original image during the sharing phase. Experimental results and analysis of the proposed scheme demonstrate a better performance than other schemes and confirm a high probability to resist brute force attack.

This document analyses the performance of subspace signal processing techniques applied to ground penetrating radar (GPR) images in order to reduce the amount of clutter and noise in the measured GPR image. Two methods considered in this work are Principal Component Analysis (PCA) and Independent Component Analysis (ICA). An approach to combine those two techniques to improve their effectiveness when applied to GPR data is proposed in this paper. The experiments performed to gather GPR data and evaluate proposed algorithms are also described. The aim of undertaken experiments is to replicate conditions found in water reservoirs where cracks and holes in the reservoir foundations and joints cause excessive water leakages and losses to water companies and the UK economy in general. Performance of implemented algorithms is discussed and compared to the results achieved by a highly skilled human - GPR image analyst.

In this paper a hybrid method for segmentation of 2-D medical images of human organs is proposed. To capture object's global properties the Markov random fields (MRFs) are involved. The result of the MRF segmentation is used to recalculate gradient vector flow based snake external forces in a further snake progression step. A MRF MAP solution is estimated using the MMD algorithm. To reduce the computation time, MRF estimation is restricted only to sites, where the state uncertainty is higher than some threshold value. Tests on MR slices and ultrasound images of the left ventricle of the human heart are reported.

In this paper, we propose a novel method to filter the keypoints and reduce redundant keypoints. SIFT (Scale Invariant Feature Transform) is one of the most robust and widely used methods for image matching and object recognition, which is robust to illumination changes, image scaling and rotation. However SIFT generates a large number of redundant keypoints in the background of the scene. Based on saliency detection and salient region selection, the keypoints out of the selected salient region are pruned in our method. The experimental results show that though the repeatability in our method is a little lower than original SIFT (less than 6%), the number of keypoints in our method is significantly reduced (more than 33%).

Based on compressed sensing, a new bit-plane image coding method was presented. Due to different important for different image bit-plane, the new method is robust to bit error, and has the advantages of simple structure and easy software and hardware implementation. Because the values of the image bit-plane are 1 or zero, one order difference matrix was chosen as sparse transform matrix, and the simulation show that it has more sparse presentations. For the general 8-bit images, its have 8 Bit-plane, eighth Bit-plane is Most Significant Bit-plane, so we can adopt more measure vectors for reconstruction image precision. At the same time, this kind of image codec scheme can meet many application demands. The method partitioned an image into 8 bit-plane, and made the orthonormal transform using the one order difference matrix for each bit plane, and then formed multiple descriptions after using random measurements of each bit plane. At decoding end, it reconstructed the original image approximately or exactly with the received bit streams by using the OMP algorithms. The proposed method can construct more descriptions with lower complexity because the process of bit plane data measuring is simple and easy to hardware realize. Experiment results show that the proposed method can reconstruction image with different precision and it can easily generate more descriptions.

In order to solve the problem of mass concurrent access to streaming media, this paper designs the application of streaming media services based on cloud computing with technologies about cloud computing. With the analysis of the actual demand, the system the paper designs includes three parts: streaming media resource center, streaming media edge node and intelligent load balance system. Streaming media resource center could manage and distribute streaming media resources; streaming media edge node is responsible for replaying requests of streaming media playing directly; intelligent load balance system would schedule system loads according to the current state of users requests automatically. After experiments, it proves that the system has good performance and practical value.

Compressed sensing is a novel signal sampling theory emerging recently. It is a theory that signals could be sampled far below the Nyquist sampling rate. This paper introduces compressed sensing theory into the application of infrared video, proposes a new residual reconstruction algorithm, and establishes a new infrared video codec model with random Gaussian matrix as the measurement matrix and with orthogonal matching pursuit algorithm as the reconstruction method. On the platform of Matlab, this paper performs the reconstruction of infrared video frames. The simulation results verify that the proposed algorithm can provide a good visual quality and speed up evidently by comparison with conventional algorithm.

Robust key point detector plays a crucial role in obtaining a good tracking feature. The main challenge in outdoor tracking is the illumination change due to various reasons such as weather fluctuation and occlusion. This paper approaches the illumination change problem by transforming the input image through colour constancy algorithm before applying the SURF detector. Masked grey world approach is chosen because of its ability to perform well under local as well as global illumination change. Every image is transformed to imitate the canonical illuminant and Gaussian distribution is used to model the global change. The simulation results show that the average number of detected key points have increased by 69.92%. Moreover, the average of improved performance cases far out weight the degradation case where the former is improved by 215.23%. The approach is suitable for tracking implementation where sudden illumination occurs frequently and robust key point detection is needed.

The camera calibration is the basis of putting the computer vision technology into practice. This paper proposes a new method based on camera calibration for diameter measurement of gear, and analyses the error from calibration and measurement. Diameter values are gained by this method, which firstly gets the intrinsic parameters and the extrinsic parameters by camera calibration, then transforms the feature points in image coordinate extracted from the image plane of gear to the 3D world coordinate, lastly computes distance between the features points. The experiment results demonstrate that the method is simple and quick, and easy to implement, highly precise, and rarely limited to the size of target.

Study all-derivable points in operator algebra. Using the operations of linear mapping and matrix algebra, and the related results of nest algebra theory, we show that the matrix (the first row and the second column element is the unit operator, the second row and the second column is invertible operator) is an all-derivable point of the second order operator matrix algebra.

In this paper, an effective slow motion replay detection method for tennis videos which contains logo transition is proposed. This method is based on the theory of color auto-correlogram and achieved by fowllowing steps: First,detect the candidate logo transition areas from the video frame sequence. Second, generate logo template. Then use color auto-correlogram for similarity matching between video frames and logo template in the candidate logo transition areas. Finally, select logo frames according to the matching results and locate the borders of slow motion accurately by using the brightness change during logo transition process. Experiment shows that, unlike previous approaches, this method has a great improvement in border locating accuracy rate, and can be used for other sports videos which have logo transition, too. In addition, as the algorithm only calculate the contents in the central area of the video frames, speed of the algorithm has been improved greatly.

A method of laser energy density distribution measurement is presented in this paper. Dot-matrix method and CCD imaging method are combined in pulse laser far-distance energy density measurement system. The laser energy data is received by detectors on the reflective board, and the laser spot reflected by the reflective board is acquired by the CCD camera. The relation of laser spot image gray and energy data is gotten by image processing algorithms, and energy density distribution is analyzed according to the model of laser energy. Algorithms principle and process are presented detailedly in this paper.

Flight safety is very important issue for aviation industries. Analyzing the flight accidents on the basis of 2-dimensional image is hardly to illustrate the complex injuries of passengers in the flight cabin. However, how to illustrate the flight accident is a challenge from 2-dimensional space to 3-dimensional space. This study proposes a particle swarm optimization approach for improving the identification of objects from 2-dimensional image. The recognition results provide the information for building 3-dimensional systems for flight accident investigators. The experiments also show that it is a feasible approach for improving the identification of image objects.

Recently several spectral-spatial classification methods had been presented and applied to pattern recognition of hyperspectral imagery. However, the present methods are especially suitable for classifying images with large spatial structures in spite of the derived classification accuracies of above 90%. To classify hyperspectral images with larger as well as smaller spatial structures, a novel spectral-spatial classification method was presented and tested on an Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) image with 145×145 pixels and 220 bands. Firstly, the AVIRIS image was implemented a spectral mixture analysis using minimum noise fraction (MNF). Based on the obtained n-dimensional eigenimage, support vector machine (SVM) was used to classify the AVIRIS image. Simultaneously, the eigenimage was calculated the mathematical morphology-based image gradients for the n dimensions so to obtain n watershed segmentation images. Finally, the SVM classification map was turned into several new ones through a series of post-processing. The experimental results verify that the proposed spectral-spatial classification method has the capability to detect larger as well as smaller spatial structures in hyperspectral imagery.

Remote telescope control is significant important for the astronomical site testing. Basing on ASCOM standard, a prototype of remote telescope control system has been implemented. In this paper, the details of the system design, both server end and client end, are introduced. We tested the prototype on a narrow-band dial-up networking and controlled a real remote telescope successfully. The result indicates that it is effective to control remote telescope and other devices with ASCOM.

In this paper, we propose a method to improve the problem that the assistant lane marks caused by pulse. We also define a method to distinguish the assistant lane marks' error rate objectively. To improve the problem, we mainly use the Sobel edge detection to replace the Canny edge detection. Also, we make use of the Gaussian filter to filter noise. Finally, we improve the ellipse ROI size in tracking part and the performance of the FPS (frame per second) from 32 to 39. In the past, we distinguished the assistant lane marks' error rate very subjectively. To avoid judging subjectively, we propose an objective method to define the assistant lane marks' error rate as a standard. We use the performance and the error rate to choose the ellipse ROI parameter.

Aims: Designing of a filter which can separate out the target from background in a given image. Methods: The original gray image is down-sampled by rejecting every alternate pixel values between two columns and two rows. A 3-step down-sampling was done to avoid major information loss. A 3-step up-sampling was done by replicating the lower row and the right column from the down-sampled data matrix to obtain the original size of the matrix. The image matrix thus obtained was subtracted from the original image. Results: The iterative down-sampling and up-sampling matrix gives the background information. Subtraction from original image obtains the target. Thus filters the background.

This paper presents a method that integrates local fractal dimension and threshold filter into a region growing algorithm for the segmentation of man-made regions, which are the region of man-made objects. First, we propose a new method of estimating the local fractal dimension of the nature image to make more stable and accurate. Then, the local fractal dimension feature is integrated into the region growing algorithm. Last, we use threshold filter method twice to produce a nearly accurate segmentation of man-made regions. The effectiveness of the proposed method whose aim is accurately segmenting man-made regions is confirmed through computer simulations.

According to the growing Web data mining present situation, this paper proposes a based on support vector machine and clustering Web mining new methods and based on support vector machine support vector won't appear in the two types of samples is outside of the division between correct theory, through the introduction of the clustering of centroid, class radius, class from the concepts, such as the center, which could well solve quickly and accurately remove the support vector problems, ensure the algorithm generalization. The experiment showed that the improved algorithm can quickly and accurately to the training sample of exclusion and has good generalization problems to solve.

Undistortion methods based on perspective invariants play an important role in computer vision. The key of these methods is how to choose an proper measure to describe the perspective invariance of undistorted image features. We propose a new measure based on the homography between the control points in undistorted image and the pattern. The less the distortion is, the less the mapping error is. A new lens distortion calibration method is also put forward which uses this measure to search for accurate distortion parameters by iterative optimization. Comparing with other proposed measures based on perspective invariants, our measure is both concise and comprehensive. Both synthetic and real experiments show that our method performs well in accuracy and runtime.