基于BP神经网络的点云去噪与空洞修补 集成算法研究

onsidering the importance of balance between accuracy and efficiency in processing of 3D laser scanning data, based on relative theory of Back Propagation neural network, an algorithm which integrates noise elimination and data repairing is put forward in this paper. Through BP network's strong ability of non-linear approximation, real surface is simulated, and measured data are replaced by predict output ones. In this process, noises are eliminated, and lost data caused by noise elimination are repaired automatically. To prove the feasibility of this method, BP network is set up through Matlab7.0 software platform and experiment is carried out base on scanning data of Guishan Han tomb. The result shows the integration algorithm, which has obvious advantages in accuracy and efficiency, is feasible.