Cthe Research And Development Of Measure System Based On Binocular Vision

The vision measurement technique based on the computer vision, as it has the features of non-contact, real-time and on-line measurement and high accuracy, has been developed quickly and used widely. The measurement based on binocular stereo vision, because of its flexibility and easy system structure, becomes a vision measurement method of great potential for development. This paper makes thorough researches and studies of the stereo vision measurement based on binocular vision. The main contents are as follows:Firstly, the measure system’s parameters are calibrated. Linear model of camera is conducted in detail, and the distortion of imaging system is analyzed systematically. Then, a high-accuracy calibrating model including the radial distortion and the tangent distortion of lens is built, which is applied to calibrating the parameters of the system.Secondly, the stereo matching is researched systematically. The paper puts forward an image dense match method, which combined the advantage of image matching based on feature and image that based on region. On the basis of thorough research of the classic SIFT (Scale Invariant Feature Transform) algorithm, the SURF feature detection algorithm is conducted in detail. Then, the paper improves the algorithm, including the utilization of the color information, epipolar geometry restraint to boost the matching accuracy. Besides, the paper adds the process of dense matching based on the matched feature points to increase the number of matching points. According to the comparison experiments, the robustness and effectiveness of the improved method are confirmed.Finally, three-dimensional reconstruction based on the point cloud is realized. The process of obtaining the object’s three-dimensional coordination is analyzed and conducted mathematically. And the point cloud is input into the IMAGEWARE to require the distance between two points. The results are compared with that obtained by real measurement, which confirms the effectiveness of the point cloud.The experiments give a believable measurement result. It is a foundation of high-precision measurement. Furthermore, it has a guiding significance for research on stereo vision measurement system.