The Three-Dimensional Reconstruction Based On Stereo Projection

Nowadays the research in three-dimensional objects recognition is very hot. With the development of science and technology, the application range has been wider and wider. Especially in recent years, the development of digital image processing technology and high performance computer coming into being provide favorable conditions for computer vision research and application.The three-dimensional reconstruction based on stereo projection is a technology to obtain three-dimensional data by using two cameras that are placed in different orientation. The position and depth of points extracted through features extraction can be obtained by the disparity and stereo matching.The integrated stereo vision system generally includes camera calibration, features extraction, stereo matching, depth recovery and depth interpo- Lation, etc. In this paper, the camera calibration,features extraction and the stereo matching are mainly discussed. First, In terms of perspective projection transformation, a non-linear camera model considering the one term radial distortion was used. The thesis uses the linear least square method to calculate the exterior and interior parameters step by step. there isn't necessary any accurate calibration target and complex operation. For image pre-processing, noise smoothness, image sharpening and image segmentation are discussed in detail. Subsequently, edge detection using differential operator are introduced. For image matching, the area-related method is used in the first step to perform the rough points and then accurate matching be perfect by using similitude constraints in two images. Algorithm is realized by program of Visual C++. last we give the steps of reconstruction and some points value.At the end of dissertation, the summary and further research are pointed out.