Research And Application Of Three-dimensional Measurement Based On Binocular Stereo Vision

Stereo binocular computer vision has always been a hot research topic in the study of computer vision technology. It collects the object image by simulating the human eye with the binocular cameras, and establishes geometric relationship among object, object image and the camera, and finally restores the three-dimensional information of the object. It can be used in all kinds of areas, such as visual navigation by intelligent robot, component processing and testing, three-dimensional measurement of objects, medical surgical navigation, networking, military, medicine, etc. Therefore, it is very significant to research the binocular stereo vision for three-dimensional measurement and reconstruction of objects. Especially the study for the accuracy of three-dimensional measurement and reconstruction is more important towards the binocular stereo vision.This article did research in three aspects, the camera calibration technique, image matching and three-dimensional coordinate’s reconstruction in the process of three-dimensional reconstruction and measurement, and the research aims to make it is faster, accurate, and more efficient to measure and reconstruct the three-dimensional object. This article proposed some relative solutions by analyzing the existing problems in current techniques. The main work is as follows:(1) The big difference between the internal and external calibration parameters makes the error generate in restoring the three-dimensional coordinates when calibrating binocular camera by plane calibration. On this issue, this article proposed a parallel axis imaging model (PAM) to rectify the calibration parameters. This model keeps two cameras’optical axis parallel by using the principle of polar constraints, thus leading to that the cameras’ imaging planes always are coplanar. Because the offset is just exit in X direction, so the internal and external parameters of the cameras could maintain consistent basically, corresponding to the principle of ideal binocular stereo imaging. The experiments show that the model can greatly improve the measurement accuracy for three-dimensional measurement of objects. At the same time, the measurement accuracy has been improved further by aligning two measurement parameters, the measuring distance and measuring height.(2) This article designed a fast image matching algorithm based on the PAM. It is proved by the result of the experiment that matching accuracy can reach above90%by using the algorithm for the similarity matching of image feature points, and compared with traditional similarity matching algorithm, the maximum speed-up ratio of the matching algorithms can reach80times. Considering the problems in practical application, in the comparison of measurement experiments, the results indicate that the precision has obviously improved, and the reduction is well work. This algorithm laid the foundation for practical applications.