Research On Visual Tracking Algorithm Based On Information Fusion

With the rapid development of computing technology and electronic technology, the technological level of the whole society has been improving. As the increasing of social security awareness, needs to fully utilize a large amounts of video data resources, and therefore the demand of visual tracking is also increasing. Visual tracking has now been widely applied in all aspects of society, and become one of the major domestic and international research works. There are still some difficulties in the current visual tracking technology, such as how to track in the difficulties efficiently and accurately. The key question is how to describe and model the target accurately. This paper summarizes previous work, extract various features of the image, and describe the target by the way of information fusion, combined with the principle of the tracking algorithm to improve the performance of tracking. The main work in this paper is as follows.In the aspect of various information of feature which describe the target in tracking, studied the typical features such as color, texture, geometric characteristics and GIST feature, HOG feature in visual tracking. For the principles and characteristics of the features, experiment the method to prepare for the establishment of the target model.In terms of the visual tracking algorithm, the paper studies the particle filter tracking algorithm and Kalman filter tracking algorithm which based on the predicted filter algorithm, analysis and discuss its theory through experiments. Through experiments, research and analysis the accuracy and validity of the algorithm target tracking in the changes in the light, fast-moving and occlusion.This paper presents a particle filter tracking algorithm based on adaptive information fusion combined with multi characterization. Give the features different fusion coefficients according the strength of ability to identify and describe the target in the tracking. Meanwhile, combine with the particle filter tracking algorithm, establish positive and negative sample library based on the similarity of a particle in the particle filter and the target template, judge the state of the target, and update and adjust the target model adaptively. Experimental results show that the effect of the proposed method is superior to similar existing algorithms and effectively ensure the robustness and accuracy of in different tracking conditions.In the algorithm, the use of OpenCV and Matlab completed work to achieve visual tracking algorithm. Designed a software based on information fusion of visual tracking system, and feature extraction and visual tracking have been programmed. The system discussed herein algorithms are applied and implemented mutual authentication with this algorithm, achieve the video moving target tracking.Finally, to achieve the algorithm, complete the implementation work of the visual tracking algorithm using OpenCV and Matlab, design the software of a visual tracking system based on information fusion, feature extraction and visual tracking have been programmed. The algorithms are applied and implemented in this system, and achieve the target tracking in the video.