Research On Real-Time Color Three-Dimensional Modeling Technology

3D(Three-Dimensional) reconstruction technology is the hot topic in the field of computer vision research, it’s main goal is to reconstruct the 3D digital model of target object, thus can be used for 3D measurement, reverse engineering, virtual reality and cultural relics protection and so on. So, this technology has has important theoretical significance and broad application prospects. Method of 3D reconstruction which based on optical principle and combined with image processing theory, have the advantage of no contact with object, low-cost devices, fast processing speed and high accuracy, and is the popular direction of 3D reconstruction research. In this paper, we study 3D reconstruction method based on structured light. Following is the main content of the paper:1, we study the imaging model of camera, and proposed a system calibration method based on auxiliary camera. we designed a cube for calibration purpose with chessboard surface of different color, the cube is fixed with the handheld devices, then imaging the cube with auxiliary camera, identify and locate the chessboard corners on the surface of the cube, thus to calculate the position of the cube. We then deduce the relations between the position of the cube and the position of the reconstructed point cloud, and got the conclusion that the offset of position of the cube can be applied directly to point cloud rectification, thereby eliminating the error brought by point cloud registration and a large number of calculations.2, based on Harris corner detection method, we proposed a corner selection and coordinate optimization method aimed at chessboard corners. After study on Harris corner detection method, a chessboard corner selection method based on run-length coding theory was proposed which can filter out chessboard corners effectively. Then we study on coordinate optimization algorithm based on symmetry difference, local window was selected near the corner and the sum of difference between the two groups of symmetrical region was calculated, then according to the sum of each window we have the optimal coordinates of target corner. The experimental results show that these methods conform to the anticipated goal that can got more accurate coordinates of corners.3, we study 3D reconstruction method based on structured light and proposed a light pattern based on De Bruijin sequence coding technology, the light pattern is binary stripe structured light, using De Bruijin sequence to encode the stripe, the structure is simple and easy to decode. The we study extract the color of points simultaneous under the condition of binary stripe structured light, by reusing the camera images, the operating steps is reduced and texture reconstruction speed is improved. We take both the left and right camera images into consider to reduce the impact of environment light and differences between left and right cameras.4, we design and implement a complete structured light 3D reconstruction system including software part and hardware part, the system uses several low-cost consumer devices, including multiple cameras and a projector, it is easy to operate and has friendly user interface. The experimental results show that the system can reconstruct three-dimensional color model of objects quickly.