Precise Point Set Registration Algorithm for Indoor Scene Reconstruction

In this paper, we propose a precise point set registration algorithm for indoor scene reconstruction. Firstly, as for the problem that sensor samples unevenly, point-to-plane metric is introduced into our model, which effectively makes use of the surface structure information of point set and improves registration result. Secondly, to handle with noises and outliers in indoor scene, correntropy is introduced. To solve our model, we propose a variant of Iterative Closest Point (ICP) algorithm, which iteratively establishes point-to-point correspondence and calculates rigid transformation parameters. After that, the simulated experiments demonstrate the accuracy and robustness of the proposed algorithm compared with other methods. Besides, it is also used to indoor scene reconstruction and the result is satisfying.