Detection Of Human Action Based On Machine Vision

With the development of the computer and video processing technology, human behavior analysis in security monitoring, advanced human-computer interaction, video conferencing, and video retrieval based on behavior and medical diagnosis has a wide range of applications. And it is currently a hot research topic on machine vision.Focusing on the application of Intelligent Security Monitoring System, this paper designs a relevant human motion analysis system. This thesis mainly aims at the human abnormal behavior detection in static background video sequence of conduct analysis. We use the algorithm of mixed Gauss background modeling for moving target detection, and extracting the moving object. After morphological processing, we can get the better image of moving targets. After which, we extract gait contour map as human behavior characteristics for identification. In this paper, we calculate the distances between the edge and center point, between the edge and the centerline, as well as the combination of centerline and center point variance. And after 2D2PCA, we do three experiments, the experimental results show that the third experiment is comparatively well.Based on the extraction of contour lines, we combine the improvement of the Hu moments and contour features together as the behavioral characteristics. We use the Hausdorff distance on the normal and abnormal behavior similarity measurement. We calculate the variance, and remove several larger variance, the result is better. In the experiment, we adopt different behavior as normal behavior, and others as abnormal behavior, the results are all better. Therefore, this method has universal applicability.