Research Of3D-Measurement Technology And Accuracy Analysis By Binocular Vision

Vision is an important way of human to get cognition and information. Computer vision using computer technology to simulate biological visual function and recognate and understand the3D scene in the objective world, then restore the3D information from the image. Its ultimate purpose is to make the computer to have the same observation such as human eyes. Binocular stereo vision is an important branch of computervision.it is obtain the3D information through images which is photograghed by two cameras at the different positions or a camera by moving to achieve the same effect.It has very important research value and wide application prospect in medical and military fields, such as production automation, measurement testing and industrial manufacturing.According to the system structure of binocular stereo vision,the paper researched the key technologies of camera calibration include stereo calibration, stereo matching and3D information acquisition,and focuses on the analysis of stereo matching algorithm based on SURF feature. Although the algorithm has better performance in time consuming, but it has not better accuracy. In this paper, the algorithm has been improved through experimental platform to analyze and compare the experimental results of the algorithm, to verify the feasibility of the algorithm.The main contents of this paper are as follows:(1) Research on the mathematical model of camera imaging, and the calibrationprinciple of binocular stereo vision and the principle of epipolar geometry. At first, we use the Zhang’s camera calibration method to calculate two cameras to get their internal and external parameters, and then make the two image epipolar line alignment through the three-dimensional correction.(2) Research on stereo matching principle and the method based on SURF operator, matching method and decision vector detection and extraction of feature points, feature generation, and the improvement of the algorithm. Combined method of using SURF operator and wavelet transform the binocular stereo vision system to achieve fast matching which also can eliminate the error matching points and reduce the false match rate. (3) Based on OpenCV database to build experimental platform of binocular stereo vision and program system software include image acquisition, camera calibration, stereo calibration, stereo matching and the other system functions.In order to verify the effectiveness of this system, design3D-dimensional information extraction experiments to show that the system has a good feasibility.