Research of Three-Dimensional Reconstruction Based on Computer Vision

Three-dimensional reconstruction is of great importance whether from the point view of visual physiology or engineering application. With the environment in practical experiments, the basic process of research is confirmed as image acquisition, camera calibration, feature extraction, stereo matching, and three-dimensional information reconstruction. In this paper, the feature point matching algorithm is prescribed in detail, including how it improved and the validity analysis of the algorithm. The tests show that the algorithm is improved not only on accuracy but also on computational efficiency. In this paper, the experimental flow of Three-dimensional reconstruction and the tests results is given. The experimental results show that when the test image pixel size to achieve a certain size, the error variation of three-dimensional reconstruction becomes smooth.