Study On Methods Of Moving Target Detection In Reverberation Environment Of Ice-covered Sea

Arctic has attracted more and more attention because of its rich resources and special geographical location.With the global warming,the activities of countries in the world in the Arctic are also becoming more frequent,and the games of the major countries around the Arctic are becoming more intense.China is a near-arctic country and it is very important to speed up the strategic layout in the Arctic region from the perspective of both economic interests and national security.The performance of active sonar is restricted by reverberation and noise.In the water area covered by ice,the strong reverberation caused by the subsurface ice scattering is the main factor limiting the performance of active sonar.How to reduce the interference of reverberation in this area is the key problem that needs to be solved to realize active detection under ice.This paper focuses on the problem of detecting a moving target under reverberation in the ice-covered sea.The recognition and model description of reverberation characteristics are the premise to explore a more effective active detection technology.Therefore,the correlation of ice-water interface reverberation in the beam space was firstly studied,and the reverberation models of gentle ice-water interface and steep ice-water interface were proposed.In the reverberation model of gentle ice-water interface,the reverberation region was divided into several sub-regions.The gentle ice-water interface reverberation model describes the reverberation process in the beamspace by describing the sub-regions' energy fluctuation.In the reverberation model of steep ice-water interface,the reverberation region was studied according to the beam direction.The reverberation model of steep ice-water interface describes the reverberation process in the beamspace by describing the fluctuation of the reverberation envelope in all directions.In this paper,the parameters and properties of the two models were discussed,the influence of the model parameters on the model properties was summarized,the ice reverberation experiments were carried out,and the applicability of the two models was analyzed.Based on studying the characteristics of reverberation,the research on methods of moving target detection in the background of reverberation was carried out.Methods besed robust principal component analysis(RPCA)and low-rank matrix approximation(LRMA)and dynamic mode decomposition(DMD)were proposed.These methods are based on the idea of matrix low-rank and sparse decomposition,using a multi-frame data joint processing method(input multi-frame data and output multi-frame results),integrating multi-frame received data into a data matrix,by separating the coherence component with the incoherent component of the matrix,the moving target and reverberation in the data are separated,and finally the decision strategy was used to detect the target.The RPCA based method abstracted the detection of moving targets in reverberation into a mathematical problem of RPCA.The alternating direction method of multiplieris is used to solve the RPCA problem,and then the detection of moving targets is realized.This method improved the RPCA problem solving algorithm according to the characteristics of sonar data and improved the stability of the detection method.The LRMA based method combined the advantages of random algorithm and convex optimization principle,and achieved target detection from the perspective of low-rank matrix approximation.The DMD based method utilized the dynamic connection between multiple frames of data,and realized the target detection by calculating and classifying the dynamic modalities of multiple frames of data.Compared with RPCA based method,LRMA based methods and DMD based methods are more suitable for real-time processing of sonar data.The two methods improve the calculation efficiency from the perspective of iterative calculation and non-iterative calculation respectively.Under the same conditions,the calculation time is one order of magnitude less than the RPCA based method.In this paper,receiver operating characteristic curve and F-Measure were used as the evaluation indexes of the performance of the detection methods.The performance of three detection methods including RPCA based method,LRMA based method and DMD based method,were evaluated.The results of numerical simulation experiments and experiments in waters covered by ice show that the three detection methods have similar performance,and can effectively detect targets under low SRRs.The research in this paper provides a new solution and concrete means for detecting a moving target under reverberation covered by ice,and lays a technical foundation for the research of Arctic detection.