Multi-feature Fusion Particle Filter Tracking Algorithm

Video-based motion tracking is a hot topic in computer vision, the integration of image processing, advanced technology and research of pattern recognition, artificial intelligence, automatic control and computer application technology and other related areas, in public safety, military guidance, traffic monitoring, and so has been more widely used. Performance moving target tracking algorithm for tracking effect has a direct impact, high precision, robust tracking algorithm has been one of the key problems in computer vision research.This thesis describes the current popular video-based moving target tracking algorithm, these algorithms set forth the basic principles, advantages and disadvantages, and then focuses on the sport more popular target tracking Mean Shift algorithm and particle filter algorithm, as well as their target tracking field Applications.Target for the use of the color characteristics of the particle weights are updated when the color distribution of similar background and objectives or target is blocked, the tracking error prone. Therefore, we propose particle filter tracking algorithm based on multi-feature fusion, the color feature and local binary pattern(LBP) texture features combine to describe the dynamic objectives, to improve the robustness of particle filter tracking algorithm.This thesis presents a fusion particle filter algorithm based on Mean Shift and adaptive multi-feature tracking algorithm. First, based on particle filter algorithm proposed multi-feature fusion, using color and texture features a combination, to facilitate more effective coarse positioning dynamic target location, and then take advantage of Mean Shift clustering effect, and improve the efficiency of particle sampling to overcome the particle degradation, adding adaptive feature in the algorithm implementation process integration, solve object occlusion problems. Under the premise of guaranteed tracking effect reduces the computational particle filter, so as to achieve the purpose of real-time tracking of moving targets.Finally, by means of object-based video summarization system particle filter fusion algorithm based on Mean Shift algorithm and adaptive multi-feature target tracking algorithm design and implementation of the system and improve the target tracking capability video summarization system.