Research On Slow Moving Target Detection And Velocity Estimation Method For Airborne Multi-channel SAR

Airborne synthetic aperture radar moving target detection technology(SARGMTI)has excellent performance in moving target detection,moving target velocity estimation and focusing and positioning imaging,and is widely used in military and civilian fields.Due to the low detection performance of single-channel SAR-GMTI due to various influences,multi-channel SAR-GMTI has developed rapidly in recent years.However,most of the current multi-channel SAR-GMTI techniques have low detection performance and velocity estimation performance in the face of slow-moving targets.Based on this,this paper mainly studies the slow-moving target detection and velocity estimation methods of airborne multi-channel SAR.First,the multi-channel SAR moving target signal model is established,on this basis,the traditional multi-channel moving target detection methods are respectively studied,a moving target detection and velocity estimation method based on interferometric amplitude and interferometric phase,and a moving target detection and velocity estimation method based on Convolutional Neural Network(CNN).The overall structure of this paper is as follows:(1)According to the airborne multi-channel SAR slant range plane geometric model,the moving target signal model is derived,and the influence of the moving target motion parameters on the moving target imaging results is analyzed.Then,two kinds of traditional multi-channel SAR-GMTI algorithms are introduced,namely Displaced Phase Center Antenna(DPCA)technology and Along-Track Interferometry(ATI)technology.For DPCA technology,the basic principles of data time domain DPCA,range Doppler domain DPCA and image domain DPCA are deduced in detail,and the feasibility of analyzing these three processing domains DPCA is verified by point target experiments.For ATI technology,the basic principle of image domain ATI is deduced in detail and analyze its detection performance through experiments.(2)Since the interferometric phase of the slow moving target is small,only using the interferometric phase to detect the moving target has a low performance,and it is difficult to accurately estimate the velocity of the detected moving target.Aiming at this problem,a moving target detection and velocity estimation method based on interferometric amplitude and interferometric phase is proposed.The method first uses the differential relationship between the interferometric amplitude and the interferometric phase to establish the interferometric amplitude-interferometric phase two-dimensional detection threshold to achieve moving target detection.For the detected moving target,the optimal weight method is used to estimate the range velocity of the moving target,this method uses the interferometric phase to obtain the distance velocity corresponding to each pixel of the moving target,and then assigns the weight according to the intensity of each pixel of the moving target,and finally the distanceto-velocity weighted summation of each pixel is obtained to obtain the distance-tovelocity of the moving target;for the azimuth velocity,use Fractional Fourier Transform(FRFT)to calculate the distance velocity.After the moving target speed estimation is completed,the moving target focus positioning can be realized.The experimental results show that the performance of the proposed algorithm is better than that of DPCA-ATI in moving target detection,and it can also estimate the twodimensional velocity of the moving target,the estimation performance of the range velocity is higher than that of the traditional strongest pixel method.(3)Since moving object detection and velocity estimation can be regarded as regression tasks,and convolutional neural networks have strong performance in solving regression problems.Based on this,a moving target detection and velocity estimation method based on CNN is proposed.The method first uses an improved complex CNN to achieve moving target detection and focused imaging.The network makes full use of the amplitude and phase information of the SAR complex image,extends the feature map of the real domain to the complex domain,and uses the complex residual block and The complex residual dense block adaptively learns the deep features of the SAR complex image,and the complex CNN after training can realize the SAR moving target detection and focus imaging;then use a Res Net18 network fused with Squeeze-andExcitation Block(SE Block)to estimate the moving target range velocity,the network can fully learn the relationship between the complex image of the moving target and the distance velocity of the moving target.The network after training can be used to estimate the distance velocity corresponding to the moving target,and finally realize the positioning and imaging of the moving target.The experimental results show that the proposed algorithm not only outperforms other algorithms in the performance of moving target detection,but also outperforms the traditional algorithm in the comprehensive performance of estimating range velocity.