关键词: 点云处理, 蚁群聚类算法, 核函数, 映射, 光顺去噪

Abstract: To avoid phenomena of excessive smoothing and the local distortion, a kernel-function-based ant colony clustering algorithm was proposed to analyze the point clouds data, which was linear in high-dimensional feature space. Curvature and normal were mapped into feature space by kernel function, using weighted distance as the similarity measure to analyze the possible noise points and local features. Interclass variance was adopted to calculate the threshold adaptively when smoothing the normal vector. Experimental results showed that the presented algorithm had significant improvements than the classical algorithms which preserved some feature information of original data.

Key words: point cloud processing, ant colony clustering algorithm, kernel function, mapping, smoothing and denoising