Based On A Single Image Targeting And Three-dimensional Reconstruction Study

Image-based reconstruction is an important research direction in computer vision. This thesis primarily works on image-based reconstruction from single image of structured scenes, which bases on the combination of computer vision and photogrammetry. The main work is summarized as follows:1. Canny algorithm is selected to detect edges after comparing several common edge detection algorithms. In order to overcome the shortcomings of consuming time and wasting storage space, Probabilistic Hough Transform is applied to detect the object line. Then the vanishing points can be determined with fitting method in order to improve the precision.2. Camera parameters are acquired based on the properties of the vanishing points. The thesis is focused on the two methods of calculating the outer azimuth elements with photogrammetry. And then the rotation matrix is determined by the more precise method.3. The integral object should be changed into a cuboid model. Relative depth has great effects on the precision of model's three parameters, so the thesis proposes a new improved algorithm to improve precision of the parameters. Then the three-dimensional coordinates of the feature points can be calculated, and the object location also can be achieved.4. The model can be shown in the virtual environment with the application of VRML. In the process of texture mapping, the original image is divided into a lot of regular parts, and they are affixed to the surface of generated model. In this way, it can reduce the degree of deformation texture and improve the reconstruction results.The result of specific examples indicates the correctness of algorithm and feasibility of system. And it has a very broad application prospects in city planning, the reconstruction and protection of Ancient architectures, image-based measurement, virtual roaming technology and other fields.