Object Tracking And Velocity Estimation In Surveillance Video

Tracking objects in complex environments is a challenging task in intelligent surveillance area. Various situations, such as background cluttering, illumination variation and occlusion, severely weaken the robustness of tracking algorithms. In this paper, we propose an approach to estimating human size in video sequences. The novel approach can improve the robustness of human tracking. Our size estimation algorithm takes advantages of the linear relationship between the human size and the corresponding location in surveillance videos. The parameters of the linear relationship are learned with the weighted linear least square method. Before training, the human blob sizes at different locations are collected by background subtraction. Experiment results show that the proposed method outperforms the standard Cam-Shift tracking algorithm and demonstrate that the proposed size estimation method gives an effective constraint for human tracking in surveillance video.Based on the study of object tracking, we also present a prototype of an algorithm for vehicle speed estimation. Different from previous approaches, our algorithm requires no road markers and fewer manual calibrations. Based on specific projection rules, we find a relation between the in-camera coordinate and the real world coordinate. A non-linear regression is employed to estimate the model parameters. This model enables us to estimate the real world position of the vehicles directly from a video sequence taken by a surveillance camera. The algorithm shows its ability to produce accurate estimations in our experiments.