Font Size: a A A

Study Of Surveillance Video Structured Based On The Independent Targets

Posted on:2016-11-24Degree:MasterType:Thesis
Country:ChinaCandidate:N K ChenFull Text:PDF
GTID:2308330464956295Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
With the development of video security, surveillance video is explosively increasing. Facing large volume surveillance video, we urgently need intelligent video-surveillance systems. One of the key problems of intelligent video-surveillance systems is how to query and analysis the large volume surveillance video efficiently. Video Structured Description(VSD) is a key technology to solve the problem. VSD is a kind of technology to convert useful information into text file which can be understood by the computer and human. Currently, most of surveillance video is obtained by static camera, at the same time, moving targets tend to be the most important and useful information in surveillance video. Therefore, how to effectively describe moving targets is an important application field. Concerning surveillance video data of static background, this paper deeply studied VSD based on independent targets. The main works and research contents are as follows:Firstly, On the basis of motion detection and blob extraction, this paper proposes an algorithm called independent targets association algorithm based on fuzzy incidence matrix, which use the overlap ratio and centroid distance to establish fuzzy incidence matrix; then, we infer the movement in different situation and formulate corresponding decision and processing strategies towards those situations.Secondly, In the case of intersection situation, we still cannot solve the problem of target association through reasoning strategy, so this paper proposes an improved Mean Shift algorithm which contains motion-based Mean shift and local motion prediction. First of all, we calculate average speed of the independent target before entering the intersection and forecast the position of the independent target in the intersection region; then, this algorithm combined motion information with Mean Shift algorithm and reduced the disturbance of background information. Thus, this algorithm improved the accuracy of the Mean Shift algorithm for target tracking.At last, On the basis of the researches above, a kind of Video Structured Description based on the independent targets was proposed, Meanwhile, in the programming environment of Matlab, we designed the system of Surveillance Video Structured based on the independent targets. This system achieved the goal of Surveillance Video Structured and query of the independent target etc functions, Thus, it verified the effectiveness of all of the proposed algorithms.
Keywords/Search Tags:smart surveillance system, Video Structured Description, target association, incidence matrix, Mean Shift
PDF Full Text Request
Related items