Research On Moving Target Detection Method Based On Low Rank For Airborne Multi-Channel Radar Systems

Airborne Synthetic Aperture Radar(SAR)can achieve high-resolution imaging of background and moving target.Thus,SAR systems are widely utilized in applications such as ocean,traffic control and battlefield surveillance.Because airborne SAR is always in the state of looking down,the radar receiver is inevitably mixed with a large amount of ground clutter.Meanwhile,due to the Doppler spread phenomenon caused by the motion of airborne platform,weak/slow targets are often drowned out by the main lobe of clutter,which seriously affects the detection performance of moving targets.In addition,in practical applications,the problem of missing space and time sampling is caused by factors such as missing array elements and data flash loss,which makes the performance of moving targets detection be further deteriorated.Therefore,based on the robust principal component analysis(RPCA)method can analyze the characteristics of samples’ deep low-rank information,it is of great significance for the moving target detection system to further study the fast method under missing space-time two-dimensional data.This paper focuses on the problems of airborne multi-channel SAR systems in practical moving target detection,such as missing array elements,data loss and poor real-time performance.The main contents of this paper are summarized as follows:1)A algorithm based on CUR-RPCA of moving target detection for the missing array elements problem is proposed.Firstly,based on the traditional separation of clutter and moving targets by the RPCA method,the equivalent constraints of the missing data matrix and complete data matrix are constructed by seeking a projection space.And the equality constraints of partial key data matrix and complete echo signal after completion are established by CUR decomposition.This algorithm improves the suppression performance of clutter under missing array elements.In addition,it realizes the completion of all missing data only by completing partial key data through the inverse operation of CUR decomposition,which makes up for the low computational efficiency of traditional matrix completion method.Subsequently,the effectiveness of CUR-RPCA approach is verified by several comparative studies and numerical simulations.2)A approach based on TSCUR-RPCA of moving target detection is proposed to solve the simultaneous missing in space-time sampling data problem.At this point,echo data will not only be missing in single dimension data.Hence,the CUR-RPCA method is still used to complete the data in space and time dimensions respectively by cascading method,which not only greatly reduces the computational efficiency of the method,but also faces the problem of completion redundancy for part of the space-time coupled data.Therefore,the proposed method improves the CUR-RPCA method and realizes the two-dimensional expansion of the data recovery process by completing the space-time two-dimension joint core matrix,namely,simultaneously data recovery is carried out on some key nodes of space-time dimension respectively.At the same time,because the space-time joint core matrix is a full rank matrix,the proposed method can also effectively alleviate traditional completion methods of high computational complexity problems caused by the time dimension of airborne SAR which is much larger than the space channel number.Finally,the effectiveness of TSCUR-RPCA algorithm is illustrated by several comparative studies and numerical simulations.3)In the case of large-scale data,limited by the load and size of the miniaturized platform and the need of fast real-time detection,it is often difficult for the signal processor mounted on the platform to realize the fast solution of multidimensional non-convex optimization problem(TSCUR-RPCA).To this end,a rapid algorithm based on AMP-RPCA of moving target detection is proposed.In this method,the approximate message passing(AMP)theory is introduced into the solution process of optimization problem in RPCA algorithm.Therefore,the complex joint optimization problem is transformed into the low-complexity fitting of multiple local observation probability density functions.So as to develop the node update rules of the approximate message passing between nodes.By solving the message of local node in parallel,the efficient clutter suppression and fast data completion are realized.Finally,comprehensive simulation and measured data confirm that the AMP-RPCA method has superior moving target detection performance and high computational efficiency.