Three-dimensional Reconstruction Techniques Based On Stereo Vision

Binocular stereo vision is the most valuable method in Three-dimensional computer vision field. Its principle is to simulate the behavior perception of human’s visual and the physiological process of human identifying the objective world, to observe the same object from two different perspectives and get a different perspective two-dimensional image according to the principle of the imaging geometry of a pixel in the location of the two corresponding image points. Because a vital topic in the stereo vision research areas in the two-dimensional image acquisition process will be subjected by many factors, for example, the light, the shape of objects and the camera internal devices, so the robustness become a vital problem in the field of stereoscopic vision research, and the robustness is also a basic question in three-dimensional reconstruction technology.The subject is studied on the basis of the design of computer vision-based three-dimensional reconstruction, and puts forward of the design of three-dimensional reconstruction of computer vision-based. The following is a brief introduction to the topics related to the key technologies and major steps. Three-dimensional reconstruction of the objective object bases on binocular stereo vision technology, the article firstly describes the basics of computer vision, including image acquisition designed by Microsoft DirectShow to complete the digital conversion of the image, and then we can obtain camera matrix by Tsai’s method that based on the RAC footwork calibration and binocular stereo vision system calibration, and correct the images according to the matrix of the camera, next preprocess the images that have been corrected. The system uses pre-treatment technology to correct the images,the image processing includes smooth and gray-scale transformation, and then obtain less noise and high quality images, use the Harris corner detection to strike useful information of the images from pixels as to match the pixel’s corresponding relationship exactly in the stereo matching stage, and then use the corresponding point, the basic matrix and camera parameters inside and outside the three-dimensional reconstruction to regain the three-dimensional information of the target.The author makes a deeply discussion and research on the stereo matching algorithm, and proposes to apply the double-density dual tree complex wavelet transform with many distinct advantages to the stereo matching, and get the disparity map that has the most absolute advantages. The series of work will play a key roll in the following work.