An Intelligent Pedestrian Tracking Algorithm Combining Interactions Between Individuals

In recent years, since the demographic dividend is diminishing gradually and the increasing of public safety problem, the development of intelligent video surveillance has important social and economic significance. It can improve efficiency and reduce the labor force. And object tracking is the key technologies of intelligent visual surveillance and the premise of target recognition and video content understanding, etc.Target tracking has been widely used in video surveillance system, intelligent human-computer interaction, robot vision navigation, intelligent transportation and behavior analysis. There is some approaches that track pedestrians based on the characteristics in image and the historical trajectory. However, such approaches only consider its past trajectory and ignore an important issue regarding human behavior. It is generally believed that people are driven by their future destination, take their environment into account and adjust their trajectories at an early stage. To improve the target performances of tracking, based on an extended social force model, a more reasonable approach is achieved by integrating Mean shift algorithm with pedestrian environment. The main research contents in this paper are as follows:(1) In order to show how environment impacts pedestrian movements in the viewpoint of force, an extended social force model is presented by considering the interaction of target and the others.(2) According to characteristics of pedestrian tracking, directional weights and speed weights are introduced to adjust the strength of the force concerning the difference of individual perspectives and relative velocities.(3) On the basis of calculating the repulsive force, initial position of target is predicted by Newton’s laws of motion, and then Mean shift is utilized to track the target position.Experiment results show that this algorithm achieves an encouraging performance in occlusion situations. The object that moves fast or changes its moving directions quickly can be robustly tracked in real time by using the proposed algorithm. As for the speed, compare the tracking performance between traditional Mean shift algorithm and Mean shift algorithm combined with least squares method and Mean shift algorithm combined with Kalman filtering, the proposed algorithm improved the speed by 62.7%, 35.9% and 61.5%.