Research On Weak Target Refinement Algorithm Based On Multi-frame Track-before-Detect Technique

With the development of target stealth performance,the traditional detect-before-track early warning system is difficult to effectively detect and stably track the above targets.Multi-frame track-before-detect technique can effectively detect and track weak targets by using the energy accumulation of multi-frame measurement information.However,with the development of the distribution and motion characteristics of threat targets towards multiple batches,high maneuverability and grouping,the existing multi-frame track-before-detect technique faces the problems of high computational complexity of multi-target detection,large number of false targets,low detection probability of maneuvering targets,poor tracking accuracy and low detection resolution of group targets.In view of the above problems,this dissertation studies the fine processing algorithm of multi-target,maneuvering targets and group targets basing on multi-frame track-before-detect technique.The research content mainly includes three parts:multi-frame track-before-detect algorithm of multi-target,multiframe track-before-detect algorithm of maneuvering targets and multi-frame track-beforedetect algorithm of group targets.The main work and contributions of this dissertation are as follows:The first part studies the multiple targets multiple frames accumulation track before detect(MFA-TBD)algorithm.By constructing the multi-sensor homology detection model,a multi-sensor MFA-TBD algorithm for measuring trace clustering is proposed.The algorithm uses multi-sensor homology detection to eliminate clutter/noise,which can greatly reduce the computational complexity of multi-target MFA-TBD algorithm and effectively suppress the number of false targets.The detection performance is equivalent to the existing centralized MFA-TBD algorithm,and the computational complexity is equivalent to that of distributed MFA-TBD.By constructing the measurement spatial clustering model,a multitarget MFA-TBD algorithm for measurement spatial clustering is proposed.The algorithm uses the spatial distribution characteristics of targets to realize the spatial division of measurement points and traces,effectively reduces the computational complexity of multitarget MFA-TBD algorithm,and can suppress the generation of false targets.Compared with the existing multi-target MFA-TBD algorithm,the detection performance of this algorithm is equivalent,and the computational complexity is greatly reduced.The second part studies the MFA-TBD algorithm of maneuvering targets.By constructing the target trajectory fitting model based on historical measurement,an MFATBD algorithm based on target trajectory estimation is proposed.The algorithm effectively improves the detection probability of weak maneuvering targets by expanding the dimension of value function accumulation and introducing dynamic correlation gate,and greatly reduces the computational complexity of the algorithm.By constructing a multi-sensor hybrid fusion framework,a maneuvering target tracking algorithm based on the hybrid fusion framework is proposed.By dividing the fusion of different performance sensors in the network,the algorithm realizes the fine fusion of the same performance sensors and the extensive fusion of different performance sensors.Compared with a single sensor,the algorithm effectively improves the accuracy and robustness of maneuvering target tracking.The third part studies the group target MFA-TBD algorithm.By constructing a multiframe detection model based on Bayesian framework,a group targets MFA-TBD algorithm based on Bayesian framework is proposed.Firstly,the algorithm uses the time-space distribution correlation of group targets to realize measurement grouping.Secondly,using maneuver detection to modify the cluster group association strategy can improve the detection performance of maneuver group targets.Compared with the existing MFA-TBD algorithm,this algorithm can effectively improve the integrity of group targets detection in both high-resolution and low-resolution sensors.By constructing the gruop targets connected graph model,a group targets clustering algorithm based on connected graph label traversal is proposed,which can effectively realize the fast clustering of multiple group targets.