Research On 3D Reconstruction Based On Binocular Stereo Vision

Binocular stereo vision is a form of machine vision by imitating the theory of human eye to realize the purpose of allowing the machine to observe and understand the world. Due to the positional relationship between special binocular stereo vision cameras, two-dimensional image reconstruction is easily carried out and a three-dimensional scene with depth information is generated. Three-dimensional reconstruction is an important issue in the field of computer vision. Reconstruction relates to knowledge about computer vision and computer graphics, widely used in robot navigation, automation, medicine, archeology, virtual reality and other fields. Three dimensional reconstruction of binocular stereo vision has two key steps: camera calibration, image feature extraction and matching. This article focuses on the study for camera calibration and feature extracting and matching. The main contents are as follows:In the camera calibration, the camera model and calibration method were summarized in this paper. the choice of Zhang Zhengyou calibration method to calibrate the camera. According to the distortion coefficient for rectified images.Image feature extracting and matching, SIFT algorithm scale, rotation invariance under different lighting conditions while the case is still better characteristics, so this study focuses on the use of algorithms SIFT image feature extraction and matching. The experimental results suggest that this algorithm can generate high precision and plenty of matching points.Dealing with the SIFT algorithm for high probability of mismatch problem, this paper has proposed a method based on the epipolar constraint of the SIFT mismatch removal method. This algorithm has image rectification module, feature extraction and matching module and mismatching points eliminating module. We use the SIFT algorithm to match the features detected in rectified images, then use epipolar constraint to remove the mismatching points, at last ues RANSAC algorithm to optimized. The experimental results show that this algorithm can ensure the accuracy and reduce the runtime.