Research On Vehicle Target Detection Method For High Resolution SAR Images With Backgrund Disturbances

Synthetic Aperture Radar is a type of high-resolution imaging radar,which uses antenna synthesis and pulse compression techniques to obtain satisfactory azimuth and range resolution.With the development of SAR technology,a massive of high-resolution SAR images can be obtained by on-orbit satellites.To extract information in these highly complicated images and develop SAR image automatic interpretation system has become a hotspot in this field.As the base of automatic SAR interpretation,target detection has a direct impact on the classification and recognition of targets,which has important research significance.With the increase of SAR image resolution,the detection of vehicle target reflects high application value,and the interference of background also increases the difficulty of target detection.The research in this papaer is focused on high-resolution SAR images with background disturbances,and the vehicle target detection problem under multi-objective and problem under multi-objective,multi-background disturbances are considered respectively.The main work and innovation are listed as follows:Background clutter statistical modeling of SAR images-The SAR imaging principle,speckle model and statistical modeling theory are introduced in this part.Some commonly used statistical models are introduced according to the image resolution and scene distribution,as well as some commonly used model parameter estimation methods.In order to assess the accuracy of different models for clutter modeling,some commonly used fitting evaluation criteria are summarized in this part.CFAR detection algorithm of Vehicle target based on background modeling sample selection and generalized gamma model-This paper presents a CFAR algorithm for vehicle targets based on background modeling sample selection and generalized gamma model.Aim to solve the problem of vehicle target detection in high resolution SAR images with multi–target disturbances,this paper presents an innovative method of background modeling sample selection based on equivalent view and K-S distance.With the introduction of parametric estimation equation based on MoLC and the analytic expression of detection threshold,a CFAR detection algorithm based on generalized gamma distribution is proposed.At last our experiments demonstrate that our method is effective in monitoring high resolution SAR images with multi–target disturbances.Target detection algorithm based on Gamma mixed model-For high resolution SAR images with multiple targets and multiple backgrounds disturbances,Gamma mixed model is presented to detect the target.Considering the K-S distance and MDL value,mixed components number K's effects on modeling accuracy and target detection performance is analyzed.The innovation of this part is presenting a method to select the optimal component number for the target detection algorithm based on Gamma-mixed modeling,which effectively solves the problem of vehicle target detection for high resolution SAR images with multi-target and multi-background disturbances.The methods proposed in this paper proved to be effective in monitoring vehicle target detection problem of high resolution SAR images with background disturbances,according to there performances on on simulated and/or real experimental data sets,with visual presentation as well as numerical evaluation.The performance of the proposed methods is verificated on limited data sets at present,more experiments should be done to prove the robustness of the methods in the future.