Abstract
Camera-based human body detection and tracking is an important research subject of computer vision, which has a widely used in the field of military and civil. In this paper, we focus on the technology of human body tracking based on improved camshaft algorithm. Firstly, we introduce some common image noise reduction algorithms. By combination the frame difference and background subtraction methods, an improved moving target detection algorithm is proposed, by which the whole region of target can be detected. Then, with the analysis of particle filtering and traditional Cam-shift algorithm, we introduce a new human body tracking method that is able to choose the target automatically due to the detection result. On the basis of the detection and tracking results, the algorithm of motion parameter estimation is analyzed. Finally, a set of human body detection and tracking experiments are designed to demonstrate the effectiveness of the proposed algorithms.