Research On Key Technique Of Digital3D Body Based On Stereo Vision

Computer vision has been one of the focus and hot areas of computer science researching, but also a very challenging field, obtaining three-dimensi-onal scene geometry information is the most basic and important thing. As a-n important branch of computer vision, in recent years, with the developme-nt of computer hardware and software technologies, binocular stereo vision has made great progress.As an important application of binocular stereo vis-ion measurement technology, three-dimensional human digitization technolo-gy (three-dimensional body measurement technology) has gradually been f-ocused on by more and more scholars and researchers. Because this techno-logy can accurately and quickly obtain the information of three-dimensiona-1human body characteristics’size, it can provide accurate and detailed dat-a for anthropology, industrial designing, garment industry, ergonomics, medi-cal, and other fields. So it has broad application value and important rese-arch significance.As the main theoretical basis for three-dimensional body digitization, several key technologies for realizing three-dimensional human digitization, such as camera calibration, stereo correction, stereo matching and matchin-g point’s three-dimensional coordinate computing are researched and analyzed. Firstly each key technology’s research status is cleared, and each method i-s compared and analyzed, the final realization method is determined, final-ly each of the key technologies’realization are experimented by powerful development tools such as Qt and OpenCV function library, experimental results and causes of errors are analyzed.In the study of stereo matching techniques, against disadvantages in existing stereo matching methods such as many factors affect matching, prone to mismatch and so on, present a stereo matching method based on corner features marked by checkerboard, this method firstly use projector to cast the checkerboard image on the surface of be measured object, so that easily identifiable artificial feature points are added, after image acquisition by binocular vision calibrated camera and image preprocessing includes distortion correction and stereo correction, corner features in the image can be detected by corner feature detection function in OpenCV libraries, then save pixel precision image coordinates of the corner features, in order to make the three-dimensional measurement had more accurate results, calculate subpixel precision image coordinates of corner feature position by OpenCV libraries fuction too. Then corner features in a pair of images are matched by the proposed stereo matching algorithm. In order to validate the method and its accuracy, firstly use the regular geometry object for corner detection experiment,then the distance-measure experiment is done,use both method presented by this paper and BM matching algorithm provided by OpenCV and compare results. It shows that the method effectively reduces the difficulty of matching, improved the matching accuracy.Finally, three-dimensional human body digitization is achieved by stereo matching and measurement method based on binocular stereo vision presented by this paper. Firstly determine human body feature points and measurement parts, then three-dimensional human body measurement is done, final experimental results and errors are analyzed exhaustively.