A new method combining HOG and Kalman filter for video-based human detection and tracking

Both detection and tracking people are challenging problems, because the human body is non-rigid and there is occasion between the body block. In general, human detection is a prerequisite for human tracking, and tracking has no effect on human detection. However, a novel approach is proposed for human detection and tracking in this paper, changing this situation. First, improved HOG is used to extract human features in the image. Second, we make the relationship between human detection and tracking closer-detection is not only the prerequisite of tracking and it also benefits from tracking. Finally, the Kalman filter is introduced into detecting and tracking people. Our experiments have demonstrated that such a method reduces detection time, and improves human detection and tracking accuracy.