Research Of Motion Pedestrian Tracking Methods Based On Particle Filter

Pedestrian tracking has been one of the hot researches in the field of computer vision,and it is also the basis of pedestrian recognition, pedestrian abnormal behavior analysis andother advanced applications. Now it has been widely used in video surveillance,human-computer interface, and intelligent transportation, military and other fields. As oneof the mathematical tools to deal with non-linear and non-Gaussian problem, particle filterhas been widely used in the field of computer vision. This article is also based on particlefilter to research the pedestrians tracking methods. The thesis research contents are asfollows:(1) A new particle filter tracking method based on HOG and color features is proposed.The proposed method computes the effective particle number of HOG and color features,and integrates the two features by the weighted average, finally complete the pedestriantracking in the particle filter framework. As a result, it solves the unstable problems thatthe traditional particle filter tracking method based on the single color feature is prone toappear the loss of tracking and offset, and improves the stability and accuracy of thetracking algorithm.(2) A pedestrian tracking method based on QMC by particle filter is proposed. Theproposed method uses the Quasi-Monte Carlo (QMC) sampling methods to get the moreuniform distribution of particles on the basis of the HOG and color features so that it canestimate the posterior probability of the target state better. It solves the prone gap andcluster between particles in the traditional particle filter which will affects the trackingaccuracy of the algorithm, even worse, cause the failure of the filter. And the experimentshows the proposed tracking method based on QMC improves the stability and accuracy ofthe tracking algorithm. (3) Pedestrians tracking system based on particle filter has completed. The systemdesigns and implements the pedestrian detection and the improved particle filter trackingalgorithm described in this article by the Intel OpenCV open source computer visionlibrary and visual studio2008development software. It completes the automaticpedestrians tracking system, and provides a solution for the further study on the pedestriantracking method based on particle filter.