Research On Pedestrian Detection And Tracking Based On Video

In recent years, with the improvement of the performance of computer hardwareand software, computer network technology and the computer vision technology,intelligent video surveillance attracted the attention of scholars and researchers.Target detection and tracking are two core technologies in intelligent videosurveillance, the study target detection and tracking method has important practicalsignificance.At present, although many researchers have proposed lots of pedestrian detectionand tracking algorithms, but many problems still exist. Most algorithms illuminationchanges, interferences occlusion, target motion changes in circumstances affecting thelarger states, often false detection occurs, missed the target lost and so on.In this paper, a fixed camera capture video images for the study, research morepedestrians within the video object detection and tracking, mainly includes thefollowing aspects:First, after study of the common image object detection and extractionalgorithms, we propose a joint edge intensity histogram features, more effectivelydescribe the characteristics of a variety of body posture changes. After the features onthis article by SVM statistical learning training, training the available pedestrianclassifier. Thereafter, the first frame of the video frame using the improved adaptivedifferential extraction of the foreground region is substantially based on thepedestrian SVM classifier means detects the first frame of the video image of thepedestrian area.Secondly, we propose a fast tracking algorithm based on the detectionsegmentation. For the extraction of the pedestrian area in the video, training Houghvoting model and segmentation model and a model train is detected by thecorresponding target in the subsequent few frames location to realize the effect ofpedestrians tracking.Finally, the experiment shows that pedestrian detection and tracking algorithm presented in this paper can effectively carry out multiple targets in the video morestable tracking and detection, and to some extent, reduce the impact of changes inillumination and pose tracking.