The Research Of 3D Reconstruction About The Skull Parts Based On Structured Light

During Robot-assisted surgery in the medical, the augmented reality technology for improving the limited vision of doctors is very helpful, there are many research institutions at home and abroad for the research in this area. In the augmented reality technology, three-dimensional reconstruction of generating virtual objects is the key. There have been related products in the current market of foreign countries, and there are also a number of research home institutes engaged in research in this area.During the reconstruction process, there are a variety of ways for collecting the data about the objects. As blood vessels and nerves are more diverse in the skull parts, and the structured light method is flexible, wide measurement range, non-contact measurement and has many other advantages, so it is selected this method in this article. Three-dimensional reconstruction using structured light is that projecting the structured light onto the skull model, capturing the modulated images by the left and right cameras, then image processing, obtaining the feature points; After image registration, the corresponding feature points are got in the two images, using the binocular stereo vision technology to realize 3D reconstruction.In this paper, a set of vertical and horizontal stripes of structure of light is projected. Gray code has the smallest error rate, combined with line shift stripe positioning, it can double the density of data collection. In this article Gray code stripe edge is extracted by Canny operator and this paper presents a new extraction algorithm to achieve sub-pixel light stripe centerlines of the line shift stripes; Then it adds the extracted lines of vertical and horizontal stripes, selecting the intersection as the feature points. According to Gray code line shift method of decoding principle, the feature points of the same stripes are determined in the left and right images. In order to improve the matching accuracy, this article combines with epipolar matching to get a set of matching feature point pair.In the process, camera calibration is a very important step because its accuracy has a direct impact on the rebuilding results. This article uses the concentric calibration board to calculate the camera scale factor, combined with the zoom lens method of three cameras for solving the main points of image plane coordinates. According to binocular stereo vision technology, this article uses the extracted image feature points to get the corresponding spatial point coordinates. While all the feature points on the same deal, it can get the point cloud of feature point pairs.Using point cloud data to reconstruct the skull model, many solutions are based on the Delaunay triangulation method. Based on the Lawson approach, this paper adopts the method of adding point by point into the triangulation, thus it forms the initial triangular mesh. On the basis, this article uses Loop subdivision on initial grid to obtain the final results. Loop subdivision making the surface is divided into four modes generate vertex control point and the edge control points. It's second-order continuous in regular grids, and it can be achieved in the first-order boundary smoothing.Under the guidance of the theory, this article carries out experimental research, projecting six Gray code stripes and four line shift stripes of the structured light, and then projecting a group of its orthogonal structured light stripes. After processing the image, edge lines of Gray code stripes and the centerlines of the line shift stripes are obtained, they are clear, to sub-pixel accuracy. The effect of the skull parts reconstruction is more ideal, and the result of Loop subdivision is more obvious, softer surface reconstruction, better smooth of the light.