The Study On 3D Moving Object Detection System Based On Optical Flow

Moving object detection is an important subject in the field of computer vision, image processing and pattern recognition, in which the three-dimensional moving object detection is a hot research topic in recent years. Its main target is to get the motion parameters of moving targets(such as position, velocity, acceleration, etc.)and trajectory, so as to further realize the processing and analysis, to understand the behavior of moving objects for the higher level tasks. As a target detection technology, optical flow estimation has the advantages of high accuracy rate and can directly obtain the target motion parameters, which has been deeply researched and developed.This paper mainly studies the three-dimensional motion object detection system based on optical flow estimation and its related technology. The thesis consists of four parts: the based technology of optical flow estimation, image preprocessing, Information acquisition based on binocular vision, three dimensional expansion of optical flow estimation and three-dimensional moving target detection. Firstly, the basic algorithm and optimization algorithm of optical flow estimation are studied by the experimental verification and comparison. Secondly, in image preprocessing part, through technology comparison and experimental verification, the right pretreatment scheme and steps are designed for this project. Then, the binocular vision technology, the camera calibration and stereo matching technology are researched, information collection are completed for this system by using the binocular vision technology. Finally, the optical flow basic constraints and classical gradient optical flow algorithm which are Horn-Schunck algorithm and Lucas-Kanade algorithm are analyzed by 3D extension, and validated by experiments.