Research On Pedestrian Tracking And Recognition By HOG

Video-based analysis of pedestrian movement is a focus of computer vision, image processing field research. It has important practical value in image processing, intelligent video surveillance, visual navigation, information retrieval, and multimedia applications.The main contents and results of this study include:1. Analysis the characteristics of the social security video surveillance, explore problems that may be encountered, introduced the hardware structure and modules of video surveillance system. Determining the design of the system software and the problems to be solved. Initially introduced HOG, and analyzes its characteristics.2. The pedestrian tracking and motion analysis algorithm is analyzed and studied, the moving object classification methods, the pedestrian detection process and related algorithms are introduced, pedestrian detection method based on the histogram of gradient direction and support vector machine current are analyzed. Lay the foundation for later study.3. Starting from the classification feature extraction and classifier construction these two key, using interest point detector for temporal and spatial distribution of the key point in the film, through quantitative to several key points the main frequency representation. The experimental findings with 3-D HOG and KPD features than using single feature can improve the overall recognition rate.4. The validation and application of classification based on HOG feature recognition method. and achieved good results.