Stereo Matching And 3D Reconstruction Based On Binocular Stereo Vision

The testing equipment based on stereovision technology can be intelligent, miniaturized, digital, network and multi-functional.stereovision technology are of non-contact, high precision, online detection, real-time analysis and control, and continuous work. It can be used in many dangerous situations, and can be widely applied in military, medical, industrial, agriculture, forestry, aerospace and scientific research field. Binocular stereovision is an important branch in stereovision field. It simulates the vision process of humanbeing directly, and can measure the three-dimensional information of the object flexibly on many conditions.The complete binocular stereovision measuring process can be divided into the following main phase:image acquisition, camera calibration, binocular vision system calibration, image matching and three-dimensional reconstruction. The processes of camera calibration, image matching and three-dimensional reconstruction are discussed in this paper, and the image matching algorithm is deeply studied. Camera calibration is the basis of geometric correction and three-dimensional reconstruction, its accuracy determines the performance of the whole binocular stereovision system. A calibration method based on plane calibration template is studied in this paper. The plane calibration object is suitable to be used in the indoor environment, and it can reduce the complexity of the following stero matching process effectively.In this paper, the local feature is selected as the matching primitive, SIFT feature descriptor is adopted in the description of the feature points for its robustness, uniqueness and fast speed, and SIFT algorithm is applied in image matching. Firstly, the scale-space is created and the extreme points in the scale space are detected; then the feature points are located accurately and the unstable feature points are eliminated; after that, the main direction parameter of each feature point is specified; finally, the descriptor of the feature point is generated and matched. Because of the complexity of the real environment, many unnecessary feature points are generated by SIFT algorithm. This results in the waste of time and space, it is not only seriously affects the real-time of the algorithm, but also leads to decrese of algorithm accuracy. In order to improve the speed and accuracy of image matching, after the feature points of the two images are extracted and descripted by SIFT algorithm, this paper adopts an improved Kd-tree algorithm to retrieve the feature points and improve the efficiency of the feature points matching; finally the matching pairs are purified by the RANSAC algorithm and more accurate feature points matching pairs are achieved. Matching experiments are conducted aiming at different kinds of three-dimensional images, the experiment results validate the efficiency of the algorithm Proposed in this paper this paper. Besides, this paper also discuss the theory and mathematical model of the binocular stereovision measuring, and the calibration method of the binocular stereovision system.The critical technologies of the binocular stereovision measuring, including the camera calibration, three-dimensional image matching and 3D coordinate calculation are studied in this paper, then the complete three-dimensional reconstruction process is accomplished by the programing and the three-dimensional cloud picture is generated to validate the correctness and feasibility of the critical technologies, which settles a solid base for the development of a complete binocular stereovision measuring system.