Research On 3-D Reconstruction Based On Complex Point Cloud

Three-dimensional reconstruction technology is one of the rising technology in recent years, which has been widely used in industry, medicine and entertainment fields. The meaning of 3D reconstruction is to reconstruct the 3D object and show the object in the most intuitive way. Point clouds obtained from different views need to be transformed to the same coordinate system when they are reconstructed. This procedure is called point cloud registration. When the inside of the 3D object is narrow, closed, poor lighting and has large repetitive structures, the traditional 3D reconstruction methods is hard to obtain ideal results. This paper proposes a 3D reconstruction method targeted for this indoor environment.Kinect is an equipment using on the research of 3D reconstruction in recent years, which has the characteristics of light, small and easy to operate. In the experiments, a large amount of point clouds data are obtained by Kinect and we pre-process the point clouds, such as sample, smoothing, et al..While using traditional ICP(Iterative Closest Point) algorithm to register the point clouds, the problem such as local optimal and structure unclosed may occur. To solve the problem, this paper proposes point cloud registration method based on the MEMS(Micro-Electro-Mechanical System). This method acquires angles information with the help of angle sensor, then calculates the initial solution of rotation transformation, thus avoids the method trapping into local optimal solution. As the point cloud registration solution based on the global optimization proposed, it can reduce the accumulated error and structural registration errors.The experiment results show that the solution presented in this paper can achieve the 3Dconstruction effectively when facing the challenging of complicated environment.