The Generation And Analysis Of The Trajectory Of Moving Objects Based On Binocular Vision

Video surveillance can monitor the moving objects in the surveilled spot and analyse the objects' behaviors in order to provide auxiliary information for higher level understanding and decision making. The analysis of moving objects is an important orientation in current video surveillance, because it provides the important basises for analysis and understanding of behavior. The analysis of moving objects is especially rich in content, including moving detection,object tracking,trajectory generation,behavior analysis and so on.The relevant technology on basis of analysis of moving objects in single camera surveillance is more comparatively developed, but there are some tough problems. For example, there is the limited surveillance range,and no depth information and so on. Hence, it is a must to make research on analysis of moving objects based on binocular vision. Video surveillance based on binocular vision uses two cameras that have the same area of surveillance. It can mark and persure the 3D information of Objects on 3D space. It can generate and analysis the trajectories of moving objects. That can make the reaults more accurate. So the problem of trajectory based on binocular vision is the foundation and one of the key technologies in binocular vision surveillance, that need the more focus.In his paper, we research under the binocular vision circumstance with two cameras are stationary. The main contributions in this paper are as follows:(1) We introduce the related key techniques and algorithms about vedio surveillance based on binocular vision, such as camera calibration, objects detection, objects tracking and 3D information persuring. In detail, we make analysis and summary on some key technologies so as to provide some valuable references to the researchers in binocular vision.(2) Put forward a objects detection algorithm using Otsu algorithm based on color. Traditional Otsu algorithm can segment gray image. But transforming color image to gray image can make gray value in a closeing range that the color value out of range. So using the color information can give the objects detection better performance, so dose undergoing interference of illumination and shadow.(3) Constructing a trajectory generation system of moving objects based on binocular vision. Zhang's method is used to calibrate two cameras at the same time to obtain the inernal and external parameters of each camera. Detecting and tracking the moving objects, extracting the centroid coordinates of the moving objects, and generating the trajectory. At last, the system can measure the depth and velocity of moving objects and analyse them.