Generative adversarial network (GAN) has become a hot research topic in the field of image processing. As an unsupervised training model, GAN has been widely used in the field of computer vision, especially in image style transfer. The purpose of the GAN is to make the generator generate a false image, and the discriminator cannot tell whether the input image is the real image or the generated image. Compared with traditional network models, GAN model has these advantages in image style transfer: GAN is composed of two different networks, and the loss function is automatically learned by playing games with each other. GAN belongs to unsupervised training and does not need to annotate the data set, which saves a lot of work. In this paper, improved GAN models related to image style migration are summarized. Firstly, the principle and method of image style transfer based on convolutional neural network are introduced. Secondly, the status, principle and prospect of GAN are introduced, and the causes of gradient disappearance and mode collapse of GAN are analyzed in detail. On this basis, the principles, advantages and disadvantages of CGAN, DCGAN, CycleGAN and StarGAN V2 network models are introduced. Finally, it summarizes the current problems and future research directions of style transfer based on GAN.
Among pathological images, there are usually some typical patient structures. For these typical images, template matching can be used for diagnosis, thereby the workload of doctors can be reduced. In this paper, pathological images of breast masses are studied in depth. A typical breast masses pathological template is established. Three template matching methods, correlation coefficient matching, correlation matching method as well as square difference matching method are used for experiments separately, and the timeliness and effectiveness of them are evaluated in terms of computing time and matching accuracy. The experimental results show that among these three template matching methods, the accuracy of correlation coefficient matching is the highest, and it is able to overcome the interference of angel rotation and noise. However, it is also consumes the longest time.
The premise of the rapid establishment of Free Space Optical Communication Based on Cat's Eye Modulating Retroreflector (CEMRR FSO) link is that the CEMRR terminal can modulate incident beam and reflect it reversely to the drive end. The influence of cat’s eye effect on CEMRR FSO link gain is a crucial research content. Focusing on the influence of cat’s eye effect on CEMRR FSO, a cat’s eye effect model under defocus will be established firstly. Next, the characteristics of cat’s eye optical system under positive and negative defocus oblique incidence will be studied. Finally, the influence of cat’s eye effect on the characteristics of reverse reflection will be verified through simulation experiments, which will provide the basis for the design of the CEMRR optical system.
Laser active imaging uses a pulsed laser to illuminate the target. Through imaging detection of the target's reflected light, more detailed information of the target can be obtained. It can be used in severe weather conditions, at night and in the light of fire or smoke to monitor the target and attitude measurement. This article introduces the working principle and research status of four active laser imaging methods, including commonly used point scanning imaging, continuous light imaging, fringe tube imaging and APD imaging.
Range-gated three dimensional imaging technology is a hotspot in recent years, because of the advantages of high spatial resolution, high range accuracy, long range, and simultaneous reflection of target reflectivity information. Based on the study of the principle of intensity-related method, this paper has carried out theoretical analysis and experimental research. The experimental system adopts the high power pulsed semiconductor laser as light source, gated ICCD as the imaging device, can realize the imaging depth and distance flexible adjustment to achieve different work mode. The imaging experiment of small imaging depth is carried out aiming at building 500m away, and 26 group images were obtained with distance step 1.5m. In this paper, the calculation method of 3D point cloud based on triangle method is analyzed, and 15m depth slice of the target 3D point cloud are obtained by using two frame images, the distance precision is better than 0.5m. The influence of signal to noise ratio, illumination uniformity and image brightness on distance accuracy are analyzed. Based on the comparison with the time-slicing method, a method for improving the linearity of point cloud is proposed.
Laser image data-based target recognition technology is one of the key technologies of laser active imaging systems. This paper discussed the status quo of 3-D imaging development at home and abroad, analyzed the current technological bottlenecks, and built a prototype of range-gated systems to obtain a set of range-gated slice images, and then constructed the 3-D images of the target by binary method and centroid method, respectively, and by constructing different numbers of slice images explored the relationship between the number of images and the reconstruction accuracy in the 3-D image reconstruction process. The experiment analyzed the impact of two algorithms, binary method and centroid method, on the results of 3-D image reconstruction. In the binary method, a comparative analysis was made on the impact of different threshold values on the results of reconstruction, where 0.1, 0.2, 0.3 and adaptive threshold values were selected for 3-D reconstruction of the slice images. In the centroid method, 15, 10, 6, 3, and 2 images were respectively used to realize 3-D reconstruction. Experimental results showed that with the same number of slice images, the accuracy of centroid method was higher than the binary algorithm, and the binary algorithm had a large dependence on the selection of threshold; with the number of slice images dwindling, the accuracy of images reconstructed by centroid method continued to reduce, and at least three slice images were required in order to obtain one 3-D image.
Good target detection and tracking technique is significantly meaningful to increase infrared target detection distance and enhance resolution capacity. For the target detection problem about infrared imagining, firstly, the basic principles of level set method and GAC model are is analyzed in great detail. Secondly, “convergent force” is added according to the defect that GAC model is stagnant outside the deep concave region and cannot reach deep concave edge to build the promoted GAC model. Lastly, the self-adaptive detection method in combination of Sobel operation and GAC model is put forward by combining the advantages that subject position of the target could be detected with Sobel operator and the continuous edge of the target could be obtained through GAC model. In order to verify the effectiveness of the model, the two groups of experiments are carried out by selecting the images under different noise effects. Besides, the comparative analysis is conducted with LBF and LIF models. The experimental result shows that target could be better locked through LIF and LBF algorithms for the slight noise effect. The accuracy of segmentation is above 0.8. However, as for the strong noise effect, the target and noise couldn’t be distinguished under the strong interference of GAC, LIF and LBF algorithms, thus lots of non-target parts are extracted during iterative process. The accuracy of segmentation is below 0.8. The accurate target position is extracted through the algorithm proposed in this paper. Besides, the accuracy of segmentation is above 0.8.
According to the working principle of laser active detection system, the paper establishes the optical target laser active detection simulation system, carry out the simulation study on the detection process and detection performance of the system. For instance, the performance model such as the laser emitting, the laser propagation in the atmosphere, the reflection of optical target, the receiver detection system, the signal processing and recognition. We focus on the analysis and modeling the relationship between the laser emitting angle and defocus amount and “cat eye” effect echo laser in the reflection of optical target. Further, in the paper some performance index such as operating range, SNR and the probability of the system have been simulated. The parameters including laser emitting parameters, the reflection of the optical target and the laser propagation in the atmosphere which make a great influence on the performance of the optical target laser active detection system. Finally, using the object-oriented software design methods, the laser active detection system with the opening type, complete function and operating platform, realizes the process simulation that the detection system detect and recognize the optical target, complete the performance simulation of each subsystem, and generate the data report and the graph. It can make the laser active detection system performance models more intuitive because of the visible simulation process. The simulation data obtained from the system provide a reference to adjust the structure of the system parameters. And it provides theoretical and technical support for the top level design of the optical target laser active detection system and performance index optimization.
Range-gated laser active imaging technology is an effective way to image detection and precise tracking of remote, dark, and small targets that overcomes the shortcomings of passive visible or infrared imaging technology, thus has important practical value and broad application prospects in the military. The paper based on the analysis of its principle, technical advantages and key technologies focus on the typical systems under atmospheric conditions at home and abroad and the latest research results, and discusses the development trends of this technology.
In order to solve the low speed and low accuracy in exacting star point which used in starlight star point navigation, this paper presents an algorithm to quickly extract the coordinates of the Navistar in the image. First of all, this algorithm extracts the coordinates of star point with a low accuracy, then extracting its diffuse plaque, in the final, get its exact coordinates. Which can reduce the amount of computation to improve navigation extraction rate while avoid the time-domain filtering of the star point of the outline and diffuse spots of gray value, solving low speed in the sky diffuse plaques star point image extraction. The experiments show that this algorithm can extract the star point while making dark star and background noise greatly reduced. At the same time, star point and diffuse plaque contour gray value can be consistent with the original image.
Based on the study of measures of the algorithm casting
defect lossless examination and characteristics of X-ray
imaging, a new automatic detection based on SURF is
presented. Firstly, the algorithm detects the interested
points of specifically component model in the standard
image samples by SURF. Then the interested points of
inspection produce are detected when the rotary
worktable makes one revolution, at the same time, the
interested points between model and produce are matched.
The number of matched points is the basis for whether
the product contains the component. Experimental results
show that this method is effective in determining the
component model well or not, which provides a novel
method for casting defect lossless examination.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.