Research And Implementing On Multitarget Detection Algorithm In Low Observable And Heavy Clutter Environment

With the development of military technology,it has been more difficult for conventional detection technology to meet the performance requirements of target detection in low observable and heavy clutter environment.Track before detect(TBD)techniques are proposed for this technical problem.It uses single-frame data without thresholded or lower threshold and retains the target information as much as possible.Then,with the continuity of the target motion,it is desired to achieve the integration of target information and estimate the target state for data of multiple frames.By constructing multi-frame log likelihood ratio(LLR)formula,maximum likelihood-based TBD techniques can obtain excellent detection and estimation performance for low observable and even very low observable targets.Maximum likelihood-based TBD techniques are divided into maximum likelihood probabilistic data association(ML-PDA)algorithm and maximum likelihood probabilistic multi-hypothesis tracking(ML-PMHT)algorithm based on the different association model between measurements and targets.Although the two algorithms have the similar detection performance in single target scenario,ML-PMHT is outstanding in multitarget or multipath scenario,which is due to the advantage of “soft” association model between measurements and targets.It allows that multiple measurements can be originated from a target and the associations between measurements and targets are mutual independent.Therefore,the focus and results of this thesis are mainly based on ML-PMHT.The main contents are:1.The single-target LLR formulas of ML-PDA algorithm and ML-PMHT algorithm are discussed,respectively.Then,the complete process,which includes search,optimization and detection threshold of LLR,is discussed.Moreover,the detection performance of the two algorithms for single target is compared and analyzed.2.Joint maximum likelihood probabilistic data association(JML-PDA)and joint maximum likelihood probabilistic multi-hypothesis tracker(JML-PMHT)are discussed,respectively.It shows the outstandingly extended ability of ML-PMHT.Then,a multitarget detection and tracking framework in a sliding-window manner is introduced.3.Considering the multipath characteristic of over-the-horizon radar(OTHR),a multipath JML-PMHT is developed for multitarget detection an tracking.The corresponding Cramer-Rao lower bound(CRLB)for the multipath scenario is derived.Moreover,by implemented in the sliding-window manner,it achieves improved estimation accuracy with less false trajectories.4.Due to the great computational burden in track maintenance of JML-PMHT,a multipath probabilistic multi-hypothesis tracker(PMHT)is proposed to improve the performance of the target state estimation,Furthermore,it effectively reduces the rate of track loss by the fusion of multipath information.