Research Of 3D Reconstruction Technology Based On Pseudorandom Coded Structured Light

Recently, with the development of technology, industry of manufacturing、civilian and medicine put forward higher requirements for 3D reconstruction technology. We developed a set of coded structured light stereo vision 3D reconstruction system, which has the advantage of no guide 、 portable 、non-contact measurement, etc., you can obtain high-precision、3D integrated data of oversized workpieces quickly.The paper discusses the generation and characterization of pseudo-random code. We designed a 162′162 pseudo-random code structured light pattern, which greatly reduces the difficulty of encoding, saves coding time and improves the accuracy of measurement. Using its 2′8 window property in any adjacent two rows, combining it with gray matching to increase the success rate of matching.We set up an experimental platform of binocular stereo vision system. We calibrate the system firstly, and then complete stereo image pair correction based on the binocular calibration results. Getting the stereo image pair of standard epipolar geometry structure. Proposing several methods to speed up the stereo matching. Using coarse-to-fine matching strategy, using epipolar constraint and the depth range of object to get the approximate location of template, according to the pyramid principle to further reduce the search ing range, and finally complete the stereo matching with the ncc gray-matching algorithm. With the triangle geometry we can rapidly reconstruct the partial surface of 3D object. We elaborate the principle of ICP algorithm and obtain integral surface of the object. We analyzed the accuracy of 3D reconstruction and registration, the absolute error of the 3D reconstruction is 0.03 mm, the relative error of the 3D reconstruction is 0.25%, the maximal error of the registration is 5mm. Finally, a series of suggestions are proposed to improve the accuracy of 3D reconstruction.