Research On UAV SAR Moving Target Detection And Imaging Method

As a microwave imaging remote sensing system for active detection,synthetic aperture radar(SAR)has the unique advantages of all-day and all-weather operation.In practical applications,it is often necessary to realize the detection and imaging of moving targets,so it is of great value and significance to study the SAR moving targets detection and imaging.With the development of aviation,electronics,information,materials and other technologies,UAV technology has made great progress.The combination of SAR and UAV is beneficial to the overall performance of UAV remote sensing system.In recent years,P-band SAR system has attracted wide attention from researchers.Compared with high-wave SAR system,P-band SAR system has stronger penetration and can detect vehicle targets traveling under shallow surface and vegetation.However,P-band SAR system needs a long synthetic aperture time to detect and image moving targets,so moving targets will appear serious defocusing and displacement problems in SAR images.Therefore,compared with high-wave range SAR system,it is more difficult to detect and image moving targets.In addition,some proposed solutions have some problems such as heavy computation,difficult real-time processing,poor imaging quality and low accuracy of parameter estimation.Therefore,the main content of this paper is the research on P-band UAV SAR system for moving target detection and imaging.The main content is divided into the following aspects:1.Aiming at the problem that P-band SAR system has a large amount of computation for moving target detection,parameter estimation and imaging,and is difficult to deal with in real time,this paper proposes a fast method for moving target detection,parameter estimation and imaging,which can focus on the target and estimate its motion parameters without using parameter search.First,the algorithm uses the second-order Keystone transform to correct the range bending,and then realizes the energy accumulation based on the range frequency domain symmetric correlation function to complete the target detection and target azimuth signal extraction.The relative velocity estimation of moving target can be realized by Phase difference and NUCPF(Non-Uniform Cubic Phase Function)processing of azimuth signal.The proposed method avoids the violent search method for parameter estimation and has strong robustness.By simulating the detection and parameter estimation effects of moving targets under different Signal-to-Noise Ratio(SNR),the simulation results show that the proposed method can also detect targets when SNR is not less than-20 d B.The parameter estimation error is less than 0.25m/s.2.Aiming at the problems of low imaging resolution and poor imaging quality of moving targets by traditional time-frequency analysis methods,a moving target imaging method based on deep learning is proposed.In this method,an Encoder-Decoder residuary block structure with skip connections is used to amplify the advantages of CNN.Different from traditional imaging methods,this method uses the powerful feature extraction capability of CNN to automatically extract the feature information of the target hidden in the defocusing curve and phase,and restores the focused image of the moving target.By avoiding the complicated parameter search,SAR moving target focusing imaging can be realized and is superior to traditional imaging methods in terms of imaging quality and efficiency.Furthermore,in order to improve the efficiency of combining shallow information with deep information and weaken the influence of defocusing of the target in shallow image on the final focusing effect,a moving target program network based on the weighted fusion of shallow information and deep information was proposed.CFAR detection method is used to detect the result images of multi-moving targets by neural network imaging.Only a low threshold can be set to realize the detection of multi-moving targets.The simulation results show that: for the verification set,image entropy,MSE,PSNR and SSIM are taken as the evaluation criteria,and the image entropy can reach0.7665,which is 50% higher than other CNN algorithms.3.P-band UAV SAR test was designed to collect measured data,and the moving target imaging method based on CNN was verified by the measured data.P-band SAR system parameters and UAV platform parameters are set.According to the set system parameters,the moving target running path is set within the beam irradiation range of P-band UAV SAR system.Finally,the test scheme is formulated.According to the test scheme,the measured data were collected.Since the UAV platform was greatly affected by its own flight jitter and environmental airflow,the traditional RD algorithm was not suitable for imaging,so the measured data was imaging by BP processing algorithm.Then the SAR image block containing the moving target is intercepted and the image processing is carried out by using the imaging method based on CNN.In the SAR image block based on CNN imaging,the moving object is refocused.Finally,the effectiveness of the CNN-based imaging method is verified by comparing with the image entropy obtained from the BP processing algorithm.