Method for thinning thick point-cloud on basis of feature sensitive projection operator

Abstract

The invention provides a method for thinning a thick point-cloud on basis of a feature sensitive projection operator. The method includes the steps of selecting point-cloud neighborhoods to establish local neighborhood information for scanning scattered point-cloud data; removing outliers, to be specific, performing point-cloud initial segmenting according to the local neighborhood information of the point-cloud to remove the outliners in original point-cloud data; performing normal direction estimation on the point-cloud, to be specific, in the local neighborhoods of the point-cloud, estimating normal direction information of the point-cloud by principal components analysis; thinning the thick point-cloud, to be specific, establishing the feature sensitive projection operator according to information of normal direction differences and distances, and iteratively updating positions of the point-cloud data to thin the thick point-cloud. The method has the advantages that the feature sensitive projection operator is provided for the scanning point-cloud data with noise, outliners and thickness, and potential feature structure of the point-cloud data is kept at the premise of the thick point-cloud thinned.

Description

Technical field [0001] The invention relates to a thickness point cloud thinning method based on a feature-sensitive projection operator. Background technique [0002] With the rapid development of 3D scanning equipment, point cloud data acquisition has become easier and easier. In the process of data acquisition, due to the influence of the surrounding environment and the limitations of the scanner itself, the acquired point cloud data will inevitably have problems such as noise, outliers, and uneven sampling. In addition, due to the complexity of the object and other reasons, it is necessary to scan the object multiple times, resulting in a certain thickness of the acquired point cloud data. Meshes directly reconstructed from point cloud data with thickness usually have problems such as non-manifold structure and multi-level surfaces, which cannot be used for subsequent data processing and analysis. [0003] The initial point cloud data processing work mainly includes denoising...

AI Technical Summary

Problems solved by technology

Although the local optimal projection operator and the weighted optimal projection operator can handle noise and outliers well, they cannot maintain the inherent characteristics of the data.

[0005] When there is a large degree of noise and thickness in the acquired point cloud data, although the existing algorithms can achieve noise removal and a certain degree of data thinning, they still cannot give reasonable output results.

Images