Extended Target Tracking Based On Probability Hypothesis Density Filter

Unlike the traditional point target,the target monitored by the high-resolution sensor in each time step is displayed in the sensor as multiple measured values:the extended target.The coordinate information,target shape,and posture information of the extended target will change with time;at the same time,the measured value of the extended target will also change due to the emergence and derivation of the target,and the unexpected situation of the detector.Extended target can provide more information,and has attracted the attention and research of a large number of scholars all over the world in the past ten years.Aiming at the problems of complex data association,explosion of calculation,and poor tracking performance of traditional extended target tracking algorithms.This paper studies the Gaussian Probability Hypothesis Density Filter(GM-PHD)based on random set theory,which avoid some data association problems of traditional target and improve the ability to track extended targets.This method is based on the Bayesian theoretical framework,and uses the transfer probability density strength instead of the traditional method which transfer the posterior probability density of the target measurement point to estimate the target state and the number and distribution of the measurement points.On this basis,the main research contents of this article are as follows:(1)When the extended targets are adjacent,the accuracy of the multi-expansion target hypothesis that the density filter divides the measurement set is often low.To solve this problem,this paper proposes a density-based clustering of applications with noise(DBSCAN)combined with a fuzzy C-means clustering algorithm(FCM)to perform two measurements on measurement points.The sub-division algorithm obtains a more accurate division result of adjacent expansion targets.At the same time,in view of the problem that the clustering algorithm of clustering by fast search and find of density peaks(CFSFDP)requires human intervention to select cluster centers,an improved method is proposed to automatically obtain cluster centers and obtain the final volume The results of the test set division are more adaptable.(2)Aiming at the problem that the traditional extended target tracking method generally uses multiple measurement points of the point target as the extended target,which is not accurate enough.In this paper,the scattering point imaging simulation of the real aircraft target model is carried out,and the scattering center of the extended target is established by graph theory.This paper proposes to use the scattering center model to obtain the tracking "aiming point" of the extended target to improve the performance of target tracking.