Design Of Pedestrian Detection And Tracking Algorithm Based On Centrist Feature

Pedestrian detection and tracking technology has a wide range of application prospects and research value in intelligent monitoring,automatic driving,robot control and many other fields.Although a lot of pedestrian detection and tracking research in the field of computer vision has been done,there are still many limitations.The frame per second(fps)of the pedestrian identification and tracking system becomes a major constraint.In order to improve the speed of pedestrian detection and enhance the real-time tracking of pedestrians,the related technology remains to be further studied.In this paper,based on the deep understanding of computer vision related principles,the pedestrian feature analysis,detection technology and target tracking technology were studied in depth.A pedestrian detection and tracking software system was designed and implemented to meet the real-time requirements.The current pedestrian identification and tracking algorithm was deeply studied in this paper.The pedestrian CENTRIST feature is extracted by a simple algorithm because the CENTRIST contour feature is implemented simply and easily,and the fast classifier is used to quickly classify pedestrian images and realize a fast pedestrian detection algorithm.Using the principle of compressive sensing,the characteristics of the tracking target is extracted.A fast pedestrian tracking module is designed and implemented,and the pedestrian image detected by the pedestrian recognition module is tracked in real time.Real-time video image is transmitted by intelligent video capture card and background program access,and for real-time video images are processed by the use of background algorithm.Finally,the pedestrian identification and tracking system is verified by testing in the actual scene.Through a large number of standard pedestrian data sets and video in realistic scenes,results show that the pedestrian identification and tracking module runs an average of 15 frames per second,and the pedestrian recognition module has a recognition speed of 20 frames per second and the true positive rate is above 75%,which satisfies the effective real-time recognition and tracking of pedestrians in video images.The algorithm can effectively reduce the target by the occlusion and loss of the algorithm on the impact of accuracy.