Abstract:

Due to the errors of instrument ranging, angle measurement, reflectivity of objects, artificial operation, occlusion and environmental changes, there are a lot of noise in the point cloud data. In order to remove noise efficiently and keep the feature information of original point cloud data, the thesis proposes a new method of point cloud denoising based on moving least square curvature feature. Firstly, the curvature information of point cloud is accurately calculated by moving least squares method, and then the point cloud is denoised according to the curvature information. Finally, the feasibility of this method is verified by the quantitative evaluation method based on information entropy theory.

Key words: curvature characteristics, moving least squaves, point cloud denoising, information entropy