Study On3D Pantograph Surface Reconstruction Method Based On Binocular Stereo Vision

Multi-view stereo vision is an interdisciplinary technology which is developed with the development of artificial intelligence, computer technology, image processing, et. Multi-view stereo vision use two or multiple images from different viewpoints for three-dimensional reconstruction or measurement of space objects, and has important research value and broad application prospects.The purpose of this thesis are applying a binocular stereo vision in three-dimensional reconstruction of the pantograph slide and preliminary exploration in using three-dimensional measurement of pantograph instead of traditional tow-dimensional image measurement. The key steps of the binocular stereo vision were studied, including the following aspects.1. Designed a binocular stereo vision hardware system for pantograph surface three-dimensional reconstruction, the system employ binocular cameras to gain binocular images and use pantograph slide as the experiment subject.2. Based on Jean-Yves Bouquet’s camera calibration toolbox, calibrated the binocular stereo vision hardware system and obtained the major intrinsic parameters and external parameters.3. A Pretreatment method of binocular images is achieved. It can restrain the noise and enhance the contrast, also enhance image edge, texture and so on detail features, used SIFT feature points matching to extract pairs of matched points, then based on these matched points calculated fundamental matrix of the binocular system, and then based on the fundamental matrix implemented epipolar lines rectifying method. In addition, considering the disadvantage of the gray area matching algorithm, presented an image segmentation method based on an improved multi-scale wavelet edge detection algorithm. The interest areas that is the pantograph slide areas in the binocular images are extracted to reduce the calculation amount of stereo matching, and to some extent, improve the matching accuracy.4. By in-depth study of binocular stereo vision matching principle, used the area gray feature to descript the corresponding features of the binocular images, and used normalized cross-correlation (NCC) to analysis similarity and obtain the disparity map.5. Completed pantograph slide three-dimensional reconstruction based on the disparity map for testing and verifying the probability of pantograph slide three-dimensional reconstruction or measurement using binocular stereo vision technology, reconstruction results achieved anticipation.