Stereo Vision Based On Bi-focus Imaging

In recent years, the technology of stereo vision including recovery of depth, reconstruction of three dimensional objects, has been the hot topic for researchers. It has been widely applied in many aspects, such as nondestructive testing, predictive maintenance, mobile robot navigation, medical image analysis, virtual reality, military survey and so on. This paper focus on research of stereo vision based on bi-focus imaging, and successfully realizes depth recovery and three-dimensional reconstruction of measure scene.The bi-focus imaging system is one kind of the monocular stereo vision for it captures stereo images for a single viewpoint with a pair of bi-focus lens. Compared with binocular measurement, monocular measurement takes the advantages of one single viewpoint, small size, strong ability of collecting images in real time and so on. This paper has a detailed description on stereo vision technology based on bi-focus imaging. Most of its research contains analysis of system, feature detecting and matching, depth recovery, the calculation of three dimensional point in world coordinates,3D reconstruction and the algorithm of disparity images。The accuracy of feature detecting and matching has a significant influence of the precision of the stereo vision for it is the basis element in depth recovery and 3D reconstruction. This paper first makes a comprehensive describe on feature extraction, and choices the detecting method on phase congruency as the one which suits bi-focus imaging system best. Then the paper analyzes the imaging system and algorithm of depth recovery, describes the contribution of system parameters, and gives the requirement of optical condition to reach the ideal precision. Based on the simulation in 3ds Max and experiments of real scene, it demonstrates that the relative systematic errors can be controlled under 2/1000 when aimed at 1000m and 400m away and errors in real scene experiment is about 2/100. Based on the result of depth recovery and the parameter of camera extracted from camera calibration, three-dimensional coordinates of point in the world space can be calculated. And by the means of triangle segment of 3D points cloud and interpolation method, the paper finally gets the 3D reconstruction of measuring region and bi-focus-binocular transform. Besides, disparity image is a method widely used to describe the depth distribution of measuring scene. The paper redefines the disparity in bi-focus imaging system and by using scaling images to evaluate disparity map, category and aggregation of points, finally calculates the disparity images of measure region from bi-focus images.