Radar Weak Target Detection Based On Refined Signal Processing

With the development and wide utilization of stealthy technology,the radar cross section(RCS)of military targets,such as aircrafts and missiles,has been greatly decreased.The reflected signals from these targets are very weak,which challenges radar detection and tracking.Therefore,based on the idea of refined signal processing,it is necessary to make full use of the useful information contained in the echo data to improve the detection performance of weak targets.The current technologies of targets detection mainly include Detect-Before-Track(DBT)and Track-Before-Detect(TBD).The TBD algorithm based on dynamic programming(DP-TBD),which belongs to TBD classical technology,is implemented easily and used widely.Therefore,the detection of weak targets will be researched in terms of the conventional DBT and DP-TBD algorithms,which are based on the idea of refined signal processing in this thesis.Simultaneously,there are parts that can be executed in parallel during the execution of the DP-TBD algorithm,which can be implemented by the multi-threading technology of the GPU and further improve the execution speed of the algorithm.The thesis will also study the implement of DP-TBD algorithm on GPU.The research contents are as follows:Firstly,in order to avoid the signal-to-noise ratio(SNR)loss caused by the traditional windowed matched filter in the pulse compression(PC)process of DBT technology,the PC processing without window matching filter is adopted to improve the detection performance of weak targets.However,the distance sidelobe of the PC output is high,which will cause the weak targets to be hidden by the adjacent strong targets during the CFAR detection process.To solve this problem,a method is proposed for the detection of weak targets in this thesis,which is based on compressed sensing(CS).Combining the refined thought,the key idea of this method is using the prior information of strong targets to find the weak targets.Then,DP-TBD based on multi-frame data joint processing is researched.The principles and mathematical model of the DP and single target DP-TBD algorithms are introduced.To overcome the drawbacks of the conventional DP-TBD algorithm,two modified methods are presented in thesis.In the presence of non-uniform clutter,which leads to performance degradation for target detection and tracking,a novel power-normalizedDP-TBD method is proposed.At low SNR,the optimal estimated track is a false track with high probability.To recognize the false track,this thesis proposes a reasonable DP-TBD method which is based on the direction statistics among adjacent frames.Furthermore,simulation results indicate that the proposed methods are superior to the conventional methods.After that,the DP-TBD algorithm is generalized to the detection and track of multiple targets.Moreover,two multi-objective extraction algorithms,including local extremum method and culling method,are respectively introduced.Finally,the physical constructure of GPU and CUDA programming are illustrated and a DP-TBD algorithm based on GPU is proposed.