Three-dimensional Reconstruction Using Optical And Image Methods

With the rapid development of the computer,3D reconstruction technology has been a research focus of computer vision and computer graphics. Three-dimensional technique is widely used in medical research, architectural design, digital city, machine vision and cultural relic protection. This paper utilizes the techniques of image processing and optics to build 3D reconstruction system of single image, and realizes these functions of color distinction, slope count and 3D reconstruction with handling single color image screened by camera.Based on VS2008, Opencv2.3.1 and Matlab7.0 software development platform, this paper studies 3D reconstruction of transparent object mainly aiming at the fixation of Monocular camera.By referring to numerous relative documents at home and abroad, this paper puts forward to the method of studying 3D reconstruction of transparent object based on concluding the pioneers experience. It was divided into four parts:color distinction, the slope count of object’s surface, rebuilding the shape of object surface, the algorithm contrast and experiment analysis. The color distinction is to distinguish the color in corresponding colorful template of the object surface color by constructing vague similar matrix; The slope count of object surface is that as there is different color turning up on different slope area on the surface of transparent object, the surface slope can be calculated by the index of color refraction; Rebuilding the shape of object surface is to achieve the height value of the object surface by Poisson equation solving the slope date. As there may exist not integral in the slope date, this paper comes up with reprocessing the slope date first and then getting the height value of object surface by Fourier transform and contrast the mistakes of 3D reconstruction root mean square by the two kinds of methods.The algorithm contrast and experiment analysis is that by the method in this paper, the result of 3D shape reconstruction by five classical methods like recovering shape by shading and mistake contrast between reconstruction height value and the real height value to analyze the shortcoming of the method of recovering shape by shading and verify the feasibility and robustness of the method in this paper.At last, I look forward to the following research after concluding all I have done and the shortcoming on the paper.