Research On Technology Of Gait Identification In Complex Scenes

Gait is a behavioral feature of dynamic evolution, efficient and fast gait recognition caneffectively improve the security of security&protection. Gait video sequence gained by thehuman detection and tracking, accuracy and high information image frame is selected forsubsequent analysis and processing, and to improve the accuracy and reliability of theprevious data of the gait recognition system. In terms of gait recognition, the paper mustconsider the requirements and condition problems of the gait recognition, to extract clear,correct gait features and design high accuracy recognition method. Characteristics-based forthe gait identification, the research content of this paper is as follows:1. The human detection technology is studied, based upon the background subtractionmethod to extract a moving target, using human feature for histogram of orientation gradient(HOG), and a linear support vector machine (SVM) classifier to detect the human area.Considering complex scenes’ human detection real-time requirements, improved Gaussianmixture model by increasing its speed of modeling, to improve the rate of human detection.2. Using the method of particle filter, to human tracking, research occlusion problem.Analyzing the problem for a long occlusion, the histogram of orientation gradient (HOG)feature combined with the particle filtering method has been improved. Judging the currenthuman target of tracking by credibility, to decide whether to carry out the goals associated,which improve the accuracy of tracking in the human body is blocked.3. In terms of gait recognition for the complex scenes, the human contour changes,caused by the problem of the low recognition rate, Use of multi-feature fusion method toimprove the recognition rate has been proposed. Consist of using enhanced gait energy image(EGEI) and static energy image to feature extraction based upon gait energy image (GEI)respectively. Using nearest neighbor (NN) classifier to identify, finally making feature fusionin decision layer, which can enhance the target recognition rate.In this paper, from the recognition of biological information and combining with therelated knowledge in biology, statistics, physics and so on, Gait recognition in complexscenes for human detection, human tracking and multi-feature fusion has been studied, whichwill provide supports for the practical application of the gait recognition system.