Research On The Improved Particle Filteralgorithm Based On Weight Optimization And Itsapplication In Target Tracking

Target tracking technology is one of the core technologies in intelligent videosurveillance, which has become a popular field of computer vision.There are broadprospects and huge practical value in security monitoring, human-computerinteraction, traffic monitoring, military guidance and other areas. Since the particlefilter has advantages in solving the problems of nonlinear non-Gaussian system. Andit is very consistent with the actual situation in the target tracking environment, andthen has been introduced to target tracking by scholars. However, there is the presenceof particle degradation problem of particle filter, then re-sampling process have beenproposed, although it solves this problem, but brought new problems-lack ofdiversity of particles. Resulted in particle impoverishment and increasedcomputational, It is particularly serious for a long time target tracking accuracy.For the above-mentioned particle impoverishment, this dissertation analyzes thebasic theory of particle filter algorithm, and researches the problems deeply, presentsa weight-based particle filter algorithm optimized by optimizing the particle weightsapproach and applies the improved algorithm to target tracking process.This dissertation describes the representation of target, target feature extractionand target tracking algorithms and related target tracking technologies for thesubsequent target tracking providing technical basis. Then discusses the theory ofparticle filter in detail, presents the improved particle filter algorithm based on weightoptimization, and proves the effectiveness of the algorithm by simulation. Finally, wedesigned the process of particle filter target algorithm based on color histogram.The experiment compared the traditional particle filter algorithm and theimproved particle filter algorithm based on weight optimization in target trackingresults, experimental results show that the improved algorithm can effectivelyimprove the robustness of target tracking.