Research On Weak Targets Detection In SAR Image Under Complex Background

Currently, target detection, as a hot research issue, has been widely used in many fields,such as military search and track system, guidance, satellite remote sensing, astronomical observation, intelligent control, etc. In SAR image, weak target is difficult to identify because of the complex background and itself scattering characteristics. Issues of weak target detection in SAR image have great significance in military and civil fields and are worth further research.This paper systemmatically investigates the issues of weak target detection under complex background in SAR image.From the needs of remote sensing image interpretation, on the basis of summarizes current technologies on target detection in SAR image, this paper analyzes the problems and shortcomings of current target detection methods. From the SAR working and imageing principles, we analyzes the characteristics of radiation, geometry and noise. On this basis, combined with SAR image characteristics of various types of targets, we analyzes the characteristics of weak target in SAR image. From the sensor factor, target scattering characteristics, background environment factor and the quantity of information of environment, we discusse the reason for target weakening. Finaly, we propose the basic idea of the weak target recognition.On the basis of analyzing SAR images background clutter characteristics, we discusse the applicability of the background modeling using central limit theorem. After a brief introduction of non-Gaussian which can be often used, we particularly investigate the parameter estimation method of a stable distribution model which does not satisfy the central limit theorem. Then we apply the generalized extreme value distribution into SAR image background modeling, and introduce the properties and parameter estimation methods. Many non-Gaussian statistical model wear used for fitting the experimental SAR image background, at last, we analyzes the capability and applicability of different models.From the basic idea of weak target detection and the characteristics of weak target, starting from the accurate modeling of the background and enhanced weak target, we systematically investigates the method of weak target. Considering weak target detection problem under complex background, choose the generalized extreme value distribution which has good modeling capability to model the relative simple background in SAR image. On this basis, we propose a CFAR target detection method based on generalized extreme value distribution. After analyzed the weakness of the existing CFAR target detection methods, form accurate modeling SAR image background, we propose an improve method for weak target detection based on wavelet multi-scale and sub-regional for homogeneous zones.Considering the characteristics of weak targets in SAR image, we studied the methods of weak target enhancement. For multiple time series images, we enhance the weak target using the method of WICA-based. On the basis of the enhanced image, the weak target can be detected effectively. For the single SAR image, we study the method based on multi-aperture coherent enhancement. On the basis of the enhanced image, the weak target also can be detected effectively. By carrying out target detection experiment with different SAR image data, we analyze and summarize the experiment effects and the existing problems.From the actual demand of radar interferometry, after analyzed the features of permanent scatterers in SAR image, summarized the problems of existing methods of permanent scatterers detection, we propose the modified detection method from the perspective of weak target detection. By carrying out permanent scatterers detection experiment with different SAR image data, we analyze the performance of our metheod and the existing problems.Finaly, considering enormous computational characteristics of target detection in radar images, combine with specific steps in our algorithm, we propose the parallel algorithm for target detection, and effectively improved the efficiency of target detection algorithms.