Research On Large-Scale Measuring System Based On Stereo Computer Vision

The measurement of large work piece is one of the most important indexes in time during the process of production. The scale of work piece is too large to measure using general measuring tools. The article do some research on the measuring system based on stereo computer vision after analyzing domestic and international research present in large-scale measurement.The system using double CCD simulate human eye to measure work piece. The measuring system is made up of image input, camera calibration, stereo matching, three-dimensional reconstruction and the result output. 2D image information through image input enter the computer, after corner detection, we can set up the relationship between the 2D image information and the 3D world information, using this relationship we can get the measuring result transformed from the 2D image information through stereo matching and 3D reconstruction. This system is based on optics, using electronic technique, computer technique, image processing technique, having many merits such as simple configuration, high efficiency and high cost performance.This article proposes a sub-pixel corner detection method based on Harris operator, after analyzing several corner detection methods, analyzes the camera calibration and its accuracy, and then proposes a stereo matching method based on corner detection. At last we reconstruct the dimension of work piece through 3D reconstruction.The experiment proves the measurement system that can attain the ideal examination accuracy, and supply the gist for the multiple vision system.