A Global Refinement Algorithm to 3D Scene Reconstruction

Point cloud registration is a central problem in many mapping and monitoring applications such as 3D model reconstruction, computer vision, autonomous driving, and others. The problem of generating maps of the environment is close to the Simultaneous Localization and Mapping (SLAM) problem. A 3D scene is often defined by a sequence of point clouds. Note that some point clouds from the sequence may not have intersections. Neighbour clouds in the sequence partially overlap and are often noisy with additive Gaussian noise. Pairwise registration finds a sequence of transformations. These geometric transformations connect neighboring clouds in a sequence. Global refinement algorithms use pairwise transformation parameters and uniformly redistribute errors using graph-based optimization. Different approaches to global refinement are based on general graph optimization, layering correction, low-rank sparse decomposition strategy, and kernel-based energy function. In this paper, we propose an algorithm to align the multiple point clouds based on an effective global refinement algorithm. The proposed method is a global refinement algorithm that evaluates rotations. For global refinement of rotations, a closed-form algorithm using matrices is used. The global refinement algorithm is non-iterative. Computer simulation results are provided to illustrate the performance of the proposed method.