Research On Dynamic Programming Based Track-before-detect Algoritmh

With the rapid development of stealth technology and the increasingly complex battlefield environment,the detection and tracking capabilities of modern radars are faced with severe challenges.Track-before-detect technology(TBD)is developed for weak targets with low signal-to-noise ratio.It breaks the traditional sequential processing method of Detect-before-track(DBT).By introducing time dimension in detect process,it obtains the detection result and the target track at the same time after processing a combination of multiple measurement frames.Thereby realizing the detect and tracking of the weak target,which is a hotspot in the research of detect and tracking technology nowadays.However,the existing TBD technology is mainly used in the infrared.In order to apply the TBD technology to the radar system,there are still many problems remaing to be solved,such as the design of merit function which has a major impact on performance of the algorithm,the real-time tracking of the algorithm,detect and tracking for different target types,detect and tracking in complex multi-target scenarios,etc.In view of the above problems,this paper conducts research both in single-target and multi-target scenarios,and the main contents are summarized as follows:(1)Analyze the differences and advantages of TBD technology compared with DBT technology.The dynamic programming(DP)theory of dynamic programming track-before-detect algorithm(DP-TBD)is presented.Then a single point target is used to model the DP-TBD algorithm and analyze the corresponding algorithm flow.(2)Aiming at the real-time tracking problem of DP-TBD algorithm and the poor detection performance for extended target.Several design methods of merit function are studied based on the single-atrget DP-TBD algorithm,which nearly determining the performance of the algorithm.The Logarithm of the Envelope Likelihood Ratio(LELR)merit function is designed to improve the detection performance of the DP-TBD algorithm for extended targets.The sliding window processing method,which compensates the real-time performance of the algorithm,and the method to avoid repeated calculation problems in sliding window processing are analyzed in detail.The real-time consecutive tracking DP-TBD algorithm for extended targets is proposed.Simulations show that the algorithm can achieve continuous tracking of extended targets.(3)As for the multi-target DP-TBD algorithm,the computational expense and the effect of the merit function aliasing effect on the performance of the algorithm are studied.Firstly,the multi-target model is established and the cause of the aliasing effect of the merit function in the multi-target scenarios is analyzed.Based on the directionality of DP integration,theoretically analysis of the method for cross-adjacent targets to achieve effective detection is given.An idea of target state transition set partition is proposed,combined with which,a parallel computing based DP-TBD method is proposed.The improvement of computational efficiency is analyzed in theory compared with the traditional multi-target loop cancellation method,Simulations shows the superiority of the proposed method towards traditional method,both in performance and computational efficiency...