Fundamental Study Of Product Configuration's 3D Reconstruction Based On Images

Inferring 3D configuration of objects from two or multiple images remains an important and difficult problem in computer vision. The significance of 3D reconstruction in many projects is not only in theory, but also in practice. A number of approaches for inferring 3D configuration have been suggested, but computer vision is still in its initial stage. The difficulty arises, of course, from the fact that two images are 2D projections of the object and they have not the adequate depth information of the object. As the development of manufacturing automation, design automation, vehicle guide, intelligent robots and facial recognition, it is imperative to propose an efficient, exact, robust, general 3D recover system.Based on the binocular stereovision, a 3D surface reconstruction algorithm has been presented and implemented. It can recover 3D surface data from two images and then reconstruct the whole 3D model from its multiple images. The theory and algorithm can be applied in recovering surface of industrial product, manufacturing automation, recovering screen of traffic accidents.The 3D object surface reconstruction system based on images(OSR_System) can reconstruct the surface of the object by binocular stereovision theory, and strengthen the reality using texture. In Chapter 3, we first propose an algorithm to recover the camera information from the multiple uncalibrated real images. With ten or more pairs of pre-picked corresponding points in the planar checkboard, the camera parameters are firstly evaluated by a linear method roughly, then optimized by our non-linear method. Experiments show that this method achieves satisfactory results. Extensively, using these resulting camera parameters, we adaptively subdivide the image and match image blocks based on the depth information coherence, then extract their depth information. With the recovered depth information, we can reconstruction the virtual images between the left image and the right image by the reconstruction algorithm in Chapter 5, and realize the object's 3D surface reconstruction.Realization of all algorithms in the dissertation use MATUVB, VC++ and integrated programming of MATLAB &VC++.