Research On 3D Measurement Technology Based On Digital Stereo Microscope

With the rapid development of computer vision technology, high precision stereo vision measurement technologies have fast expanded on electronics,mechanical, biological, material areas. Digital stereo light microscope which is treated to be a non-contact, flexible and convenient stereo vision measurement tool has been widely applied to stereo measurement, micro-assembly and micro-operation. The shortage of image features amount on object and the accuracy of stereo match will have a great influence on the measurement result during stereo measurement progress.The main work in this paper researches the three-dimensional measurement technology on stereo light microscope, focus on optimizing and improving error match pairs removal algorithm and matching interpolation algorithm, and the details are as follow:Firstly, we briefly summarize various mathematical models and fundamental theories which to be core technologies consist of camera calibration,stereo match as well as three-dimensional reconstruction. In addition, several solutions will be further put forward to optimize those key technologies which influence three-dimensional measurement result.Secondly, we deeply analyze the reason for emergence of error matching pairs in stereo match progress and present a stereo mismatching removal way,the main idea is to project the original match pairs on images by using camera calibration parameters, and then generating projection vector set between projection match pairs and original match pairs, thus eliminating abnormal projection vectors could be considered as the progress of removing mismatch pairs.Thirdly, in order to deal with the problem on sparse feature points which owe to be lack of abundant image information in calculating three-dimensional spatial coordinates, we introduce a matching interpolation method based on match-interpolation bases, and obtain more three-dimensional coordinates with the aid of matching interpolation algorithm on original image. Based on the leftimage, first of all, the image coordinates which need interpolated should be explicitly specified. After that corresponding image coordinates could be calculated according to correct match pairs existed, besides those correct match pairs would be further optimized to acquire more exact matching interpolation result.Finally, the surface of the object to be measured can be reconstruction by three-dimensional coordinates, and then analyzing the data according to the images in different conditions. At last, the result of experiment indicates that the measurement data presented on this paper truly reappears the three-dimensional information of the object.