Research On The Detecting And Tracking Of Moving Human Body Based On Monocular Vision Robot

Visual analysis of moving human motion is one of the most important research topics in the domain of computer vision. It is also one of the sign of robot intelligence. The detection and tracking of moving person aims at attempting to detect, identify and track human bodies in real time or image sequences, and more generally, to understand and describe their behaviors. Thereinto, motion detecting, identifying and tracking belong to the low-level vision problems, and behaviors understanding and describing belong to the high-level vision problems. In the thesis, the low-level vision problems including moving person detecting and tracking based on monocular vision are discussed. Our work concentrates on detecting and identifying the moving human body in the indoor environments, and tracking it using Pioneer 3-DX Robot.In the thesis, first the research background, meaning, typical applications and developments are introduced. Then the structure of the Pioneer 3-DX Robot is given, next the two sensors monocular camera and laser ranger finder that loaded on the robot and their principia are introduced. The main work of this thesis is the detection algorithm of the moving object and the real-time detecting and tracking of moving human body in indoor environment.Better tracking results are the important goal in the thesis, so the work mainly includes the detection algorithm, the detection and identification of human body and the real-time obstacle avoidance and tracking. In the detection algorithm, the algorithm based on frame difference and region grow is used, which can effectively compute the tracking region. In the identification, the features that the ratio between height and breadth of human body, the ratio between height of head and height of human body, as well as the area knowledge of human body are applied to identify the human body in the indoor environment. In the end, in order to track in real time, the perception-behavioral strategy based on laser range finder is used to avoid obstacles, while the turning angles of the camera and the motion control of robot are also given. The results of tracking show that the robot can track the human body effectively in real time.