Research On Adaptive Target Tracking Algorithm Based On Particle Filter

With the increasing level of information technology in society,the traditional industries began to use information technology to improve production efficiency and reduce human consumption,and computer vision technology has been increasingly applied to the civilian areas and military fields,biometric identification,intelligent monitoring,Unmanned,intelligent weapons and other emerging concepts began to heat up.Among them,the video target tracking technology is a classic research topic in the field of computer vision,but because of the actual scene often there are changes in light,motion state mutation,target occlusion,similar target interference and other complex situations,the current target tracking technology is still difficult To meet the needs of practical applications.The target tracking problem can be thought of as predicting the spatial position in the subsequent video sequence by the location previously known by the target of interest.This is a process of estimating and validating the current state based on a priori conditions.The idea of the state of the leaves to solve the problem.This paper is the most classic particle filter algorithm for the study,to explore the video target tracking some of the key issues,the main innovative work and research results include the following aspects:1.Aiming at the problem of insufficient diversity of particles and susceptibility to scene interference in traditional particle filter target tracking method,an improved immune particle filter target tracking method is proposed.Based on the idea of artificial immune algorithm,this method is based on the critical problems in target tracking,and adds the process of antibody memory and particle set reliability to improve the robustness of the algorithm in more complex scenes.2.The establishment of a reasonable target model is an important prerequisite for the results of particle set updating to tend to the true value of the target state.In this paper,the appearance model and the motion of the adaptive learning mechanism are proposed in view of the poor adaptability of the single target model in the traditional algorithm Model,and uses the idea of feature segmentation and background weight,and gives the corresponding likelihood calculation method.3.Aiming at the problem that the single target particle filter tracking method is applied directly to the multi-target tracking problem,a fast interactive target decision and matching algorithm is proposed.This method is suitable for the tracking method under the particle filter framework,which can improve the accuracy of multi-target tracking to a certain extent.This paper attempts to improve the performance of the traditional particle filter target tracking algorithm in the more complex real scene.The comparison experiment and analysis of multi-segment typical video are selected in the dynamic scene of Visual Tracker Benchmark test library,PETS 2009 Benchmark Data test library and car camera.Through L1-deviation,target area coverage ratio,multi-target tracking accuracy.The results show that the proposed algorithm has a better performance than the traditional one,and achieves good adaptability,robustness and real-time performance in the actual scene.