| With the rapid advancement and development of computer technology, in order to make3D graphics allow for more efficient, more accurate and even more diversified processing, the science called computer vision was born.In computer vision, binocular stereo vision is a very important branch, which obtains the2D image information of object by taking pictures from different angles through a pair of calibrated cameras based on the principle of parallax, then the2D information will received a series of procession such as stereo matching in computer and produces3D data at last. Since the system of binocular stereo vision is closest to that of humans, coupled with its advantages such as strong environmental adaptability, flexible implemental means and low cost, etc, which is widely used from scientific research to industry field.The technology in binocular stereo vision system that this paper work on is3D reconstruction, which restores object’s surface structure in computer through its point cloud obtained from stereo matching. The main research works are as follows:(1)A relatively new surface reconstruction algorithm based on implicit function called Poisson surface reconstruction algorithm has been researched, also analyzes the coping solution for the condition when sampling points from measured object are unevenly distributed.(2)A relatively new surface reconstruction algorithm based on Voronoi Diagram called Powercrust surface reconstruction algorithm has been researched, and furtherly, after a deep analysis in principle and time complexity of the algorithm, a fast Powercrust algorithm that reduces computing time while guarantees certain reconstruction accuracy has been presented.(3)Through programming, realizes the surface reconstruction algorithm of Poisson, Powercrust and fast Powercrust, then comprehensively compares the time, face count and effects of reconstruction for the three algorithm by adjusting their main parameters, and at last summarize their respective features. |
PDF Full Text Request