Pedestrian Hog And Particle Filter-based Object Detection And Tracking

It is a hot topic that research human detection and tracking of a robot in computer vision nowadays. Human detection and tracking has good application prospects and can be used in intelligent transportation, security monitoring, public safety, intelligent robotic and so on.This dissertation was based on the theories of image Processing and pattern recognition, such as HOG (Histograms of Oriented Gradients), Color Histogram, SVM (Support Vector Machine) classifier, Particle Filter, and so on. The method of pedestrian detection and tracking based on HOG and Particle Filter was proposed, and aiming at producing image sequence by the monocular vision, can effectively detect and track human targets in the scene.Firstly, the creative process for the HOG feature vector set was illustrated, including calculative gradients and vote of cell gradients, as well as the normalizing for the block. Then introduced the SVM classifier training set and the training process. Using Visual C++design HOG & SVM human detector, and successfully distinguish between human and non-human.At aspect of pedestrian tracking, the method was presented on the basis of color histogram combined with particle filter. Based on the human detection, select the human object of tracked. First of all, the samples were gathered by the particle filter so that pedestrian propagation model could be built. Hereafter, likelihood model was built, based on the color histogram in target region and particle region. Updating the template, the pedestrian could be tracked by the frames, whose coordinates of centroid in the scene could be output. Compared with the kalman filter tracker, experimental results show that the particle filter tracker processing time is short, and has better performance.