An Efficient Feature Point Extraction Algorithm for Noisy Point Clouds

Point feature extraction is an important step for point cloud data processing, which includes denoising, matching, segmentation and recognition. The quality of point cloud feature extraction has a significant impact on the outcomes of subsequent point cloud data processing. This paper proposes a point cloud feature extraction algorithm that combines the Smooth Shrink Index (SSI)[11] and Point Density Index (PDI). Voxelization is also used to speed up the algorithm. More specifically, the proposed algorithm is divided into two stages: first, a density evaluation is used to quickly filter most non-feature points, and then a combined feature extraction function is defined that takes into account both SSI and PDI and is used to identify the final feature points among the candidate feature points after the filtering step. The experimental results show that the proposed method has a good anti-noise ability and can extract feature points more completely than three commonly used methods, namely PCA [13], SSI[11], and the method in[17]. Although the feature evaluation function is partially based on the SSI method, the proposed algorithm is 30-40% faster and more correct feature points can be extracted than the SSI method. © 2023 CAD Solutions, LLC, http://www.cad-journal.net.