Research On Moving Target Intelligent Detection And Tracking Technology Of Airborne Video SAR

Synthetic aperture radar(SAR)plays an important role in microwave remote sensing.However,the traditional SAR has shortcomings in the process of ground moving target detection and tracking.In recent years,as a new radar system,Video SAR has become a research hotspot.Video SAR can not only perform high-resolution imaging,but also form high frame rate sequential images,which can image in video mode and continuously monitor the target area.Video SAR has great advantages in moving target detection and tracking direction.Video SAR detects the shadow left by moving target.Low scattering area and clutter in Video SAR image will increase the obstacle to shadow detection.In addition,in the process of tracking,the defocusing energy of the target moves irregularly in the image,which makes the traditional tracking methods unable to track the target smoothly.In this paper,we propose a new detection method for moving target detection,and combine the deep learning algorithm with target tracking.The main work of this paper is as follows:1.This paper analyzes the resolution and frame rate of spotlight mode Video SAR,describes the principle of circle trace mode Video SAR and BP algorithm,and uses step frequency signal to simulate high-resolution imaging of video SAR based on BP algorithm.In addition,the principle of moving target shadow generation in Video SAR is studied,and the characteristics of static target shadow and moving target shadow are analyzed,and the simulation is carried out by using reference image.2.Aiming at the problem of moving target detection in Video SAR,a new method of moving target detection based on Gaussian mixture model is proposed.The method uses the shadow left by the moving object in its real position for detection.Using the Video SAR imaging result data provided by Sandia laboratory,some image frames are intercepted,and the moving objects in the image are successfully detected through preprocessing,threshold segmentation,Gaussian mixture model,morphological processing and other steps.Several filtering algorithms that affect the detection are compared and analyzed.3.Aiming at the problem of moving target tracking in Video SAR,depth learning algorithm is combined with target tracking.The moving target detection method based on Gaussian mixture model(GMM)is combined with manual method to label the moving target in Video SAR data.Through the research of Re3 network,the experimental data are sorted out,the network is trained,and the influence of using different loss functions on the results is analyzed.The test set is used to test the training results,and the reasons for the poor effect are analyzed.Finally,by comparing with the MDNet network training results,the feasibility of the algorithm is verified,and the moving target can be tracked smoothly.