Pedestrian Detection And Tracking Based On HOG And Particle Filter

The detection and tracking of objects in the video is the key technology to achieve intelligent monitoring systems. It can be used in driver assistance systems, intelligent transportation, public safety and other areas. As the detection and tracking of pedestrian is an important part of the object detection, its complexity and difficulty could be increased with the variability of pedestrian posture, all kinds of interference and the uncertainty in real scene. Due to a wide range of commercial value and application prospect, the detection and tracking of pedestrian becomes a forefront direction in computer vision areas and a hot research area.In this paper, theoretical knowledge of histogram of oriented gradient features, support vector machine classifier, particle filtering is used for the research on the detection and tracking of pedestrian, and pedestrian detection and tracking algorithm based on HOG and particle filter is proposed. It can effectively detect and track pedestrian target. The main research work and results are as follows:1. When detecting the pedestrian, combination of the HOG features and SVM classifier can achieve classification detection between pedestrians and non-human, and the results are better. The method is to calculate the gradient and the pedestrian target feature extraction in the gradient direction, and then use SVM training a SVM classifier. According to the experimental results, the trained SVM classifier has strong universality.2. When tracking the pedestrian, the trained SVM classifier is applied to the tracking, initializing the detection target and detecting the pedestrian target every 5 frames. In the next video frame particle filter is used for tracking. In order to obtain the new position of the target, we should calculate the similarity by comparing the color histogram of particle area with the color histogram of the target area Because of the particle filter is robust, the tracking results are better.