Low-altitude Small Target Detection And Tracking In Low SNR

With the development of unmanned aerial vehicle(UAV),traditional target detection technology is difficult to meet the detection performance requirements of low altitude and weak targets.Track Before Detect(TBD)technology accumulates target information in multi-frame measurement data,which is an effective means to deal with this problem.Among them,the TBD algorithm based on maximum likelihood has better detection performance under the condition of low signal-to-noise ratio.The research in this thesis is mainly based on the Maximum Likelihood Probabilistic Multi-Hypothesis Tracker(ML-PMHT)framework,and the relevant theories and algorithms for urban low-altitude UAV target detection are developed,focusing on multisensor information utilization and Multipath information uses two aspects to propose enhanced algorithms and perform performance analysis.Its main research contents include:1.The TBD algorithm based on maximum likelihood is studied.Derived the loglikelihood ratio(LLR)formula under two different association models,discussed several commonly used global search methods and local optimization methods,and studied two methods for determining the verification threshold,Finally,the detection performance of the ML-PMHT algorithm is verified by simulation.2.The ML-PMHT algorithm framework based on multi-sensor is studied.The LLR formula of the multi-sensor ML-PMHT algorithm is derived,and the efficient threshold calculation method of the multi-sensor ML-PMHT algorithm and the Cramer lower bound(CRLB)formula of target state estimation error in the multi-sensor ML-PMHT algorithm are proposed.Simulation experiments show that the multi-sensor ML-PMHT algorithm improves weak target detection and tracking performance.3.The ML-PMHT algorithm using multipath measurement is studied.It deduces the LLR formula of multipath ML-PMHT and the efficient threshold calculation method of multipath ML-PMHT.A method of weak target detection and tracking in the scenario where the position of the multipath reflection point is uncertain is proposed,and the posterior CRLB(PCRLB)of target state error during tracking is derived.Simulation experiments show that through effective multi-path target information fusion and accumulation,the detection capability of weak targets and the accuracy of state estimation are improved,and the number of false tracks is reduced.