Matching Technology For Object Test Based On Stereo Vision

The research based on stereo vision is an important field in the subject of computer vision, which contributes to the achievement of three-dimensional non-contact measurement, and benefits production automation. In this paper, binocular stereo vision for the detection of the CCD camera calibration, feature point extraction and matching feature points have been studied in detail.In the part of the CCD camera calibration, an improved algorithm of the calibration method for non-linear-Tsai's two-step algorithm has been put forward, completed the calibration for binocular vision system of the single-CCD camera, and significantly reduced every error on the calibration of the CCD camera. In the part of feature points extraction, the Harris corner extraction algorithms and SUSAN corner extraction algorithms have been analyzed , the feature points extraction algorithms of Scale Invariant Feature Transform has been studied in detail, and a large number of experiments show that SIFT feature points extraction has high detection efficiency. In the part of feature points matching, the four key technologies (feature space,similar to the measure,search space and search strategy) which decide the stereo matching algorithm in the performance of the pros and cons have been analyzed. In the part of feature point matching, the refined corners matching in Harris and the SIFT feature points matching have been completed, and the experiment results show that SIFT matching algorithm is better than refined corners matching algorithm in Harris.In this paper, the system software is based on Visual C + + development platform, integrates with image processing OpenCV library function, constructs a clear logic and the stability of the work of the software framework, which meets the measuring system's requirements.