Research On Gait Recognition Based On Human Geometrical Features And Its Application Platform Design

In recent years,terrorist attacks and criminal activities are taking place all over the world.So more and more attention has been paid to public security in various countries.As a result,video surveillance has become a hotspot recently.Because people are the most important subjects in video surveillance,it is very important to identify and track the same person in different surveillance cameras which makes person re-identification become a very important direction of research on video surveillance.The recognition based on the biometrics becomes the research focus of person re-identification at present because of the uniqueness of its features.And gait recognition can play an increasingly important role with the advantages of long-distance observation and convenient acquisition compared to other biometric features(such as human face fingerprints),and it will have a very broad application prospect.In this dissertation,a gait recognition algorithm based on human geometrical features is proposed,and an intelligent video monitoring platform is set up for practical application.The main contributions of this dissertation are as follows:(1)The vanishing point extraction and the error correction of the height and width measurement is introduced to the basic algorithm of the height measurement of a target.Based on those,an automatic measurement algorithm based on monocular vision for target's height and width is designed and applied in real scene.The experimental results show that the percent error of height measurement is about 1%while the percent error of width measurement is about 3%.The result is very accurate.(2)A pedestrian recognition method in lateral view based on human geometrical features is proposed.The human geometrical features include height,step width and joint angle.With the improvement of extraction algorithm of human's lower body,the accuracy of joints,position is improved.And a classifier based on metric learning is adopted to solve the problem of large intra-class difference in gait recognition.The CCR of this method on the clothes set of CASIA is 56.5%,beyond the existing algorithms,including CGI,bSHI.(3)An intelligent video monitoring platform is designed based on the video surveillance network built in the laboratory,which realizes the functions of video playback,video content screening and gait recognition.The realization of the intelligent video analysis platform can greatly reduce the time of manually monitoring the video,enhance the storage capacity of the system,and make the pedestrian recognition process more automatic.At the same time,the platform is developed in a modular way,and the algorithm or function can be added to the platform seamlessly and can be used for practical application.