The Research Of Computer Vision Three-dimensional Reconstruction Based On OpenCV

Collaboration with the human eye and brain makes it possible to obtain, process and understand visual information. In fact the human environment information obtained about 80% is get by the visual. In recent years, with the rapid development of computer technology and digital signal processing technology, people learned to use the camera for the environment and convert images into digital signals, and using the computer to realize the visual information processing. General speaking computer vision means a computer realization of human visual function——the objective world of three-dimensional scene perception, recognition and understanding, which means that the goal of computer vision technology research is to make a computer that has two-dimensional image recognition by three-dimensional environmental information. Three-dimensional reconstruction is a technology of use the images or other information and quantitative methods to research in kind or the three-dimensional environmental information of the scene. Currently, the research result of the three-dimensional is widely used, including medical imaging, architecture, animation, action recognition and robot navigation and other fields. So it is a topic concern of computer vision in recent years.This article starts with the basic theory of computer vision, and thus leads to the research topic of three-dimensional reconstruction. Reconstruction includes camera calibration, stereo matching and three-dimensional reconstruction of three major components. Camera calibration is mainly used to get the inside and outside parameters of the used camera, preparing for definite of the camera projection matrix. In this paper, is based on linear and nonlinear combination of calibration method and completed in Matlab environment, using the toolkit for the Jean-Yves Bouguet the Matlab camera calibration toolbox. Three-dimensional reconstruction of stereo matching is the most critical step, and also the focus of this study. With the actual situation we use the sparse point-based window matching algorithm; the algorithm is mainly the image of some strong matching feature points, such as corner points, contour points. After match over, interpolation of these points to get a more complete matches. Finally, after amend the image we started three-dimensional reconstruction, the main research method is through the camera's imaging model to calculate the coordinates of three points.Demonstrated in the experimental stage, we use VC++ and Inter provided specifically for the development of the OpenCV computer vision library, the realization of the actual shooting of the three-dimensional reconstruction of objects to verify the feasibility of the theory described in this article.