摘要
针对点云数据中噪声点的剔除问题,提出了一种基于改进DBSCAN(density-based spatial clustering of applications with noise)算法的多尺度点云去噪方法。应用统计滤波对孤立离群点进行预筛选,去除点云中的大尺度噪声;对DBSCAN算法进行优化,减少算法时间复杂度和实现参数的自适应调整,以此将点云分为正常簇、疑似簇及异常簇,并立即去除异常簇;利用距离共识评估法对疑似簇进行精细判定,通过计算疑似点与其最近的正常点拟合表面之间的距离,判定其是否为异常,有效保持了数据的关键特征和模型敏感度。利用该方法对两个船体分段点云进行去噪,并与其他去噪算法进行对比,结果表明,该方法在去噪效率和特征保持方面具有优势,精确地保留了点云数据的几何特性。
A denoising method based on the improved DBSCAN(density-based spatial clustering of applications with noise)algorithm is proposed to address the problem of removing noise points in point cloud data.The statistical filtering method is applied to pre-screen isolated outliers and remove largescale noise from the point cloud.The DBSCAN algorithm is optimized to reduce computational time complexity and achieve adaptive parameter adjustment,thereby dividing the point cloud into normal clusters,suspected clusters and abnormal clusters,and immediately removing abnormal clusters.Distance consensus assessment is applied,and suspect clusters are further evaluated.By calculating the distance between the suspected point and its nearest normal point fitting surface,it is determined whether the suspected point is abnormal,effectively maintaining the key features of the data and model sensitivity.This approach outperforms other algorithms in denoising efficiency and feature retention by being implemented on hull point clouds,which ensures the integrity of the point cloud data's geometric properties.
作者
葛程鹏
赵东
王蕊
马庆华
Ge Chengpeng;Zhao Dong;Wang Rui;Ma Qinghua(School of Naval Architecture&Ocean Engineering,Jiangsu University of Science&Technology,Zhenjiang 212003,China;Jiangsu Hantong Wing Heavy Industry Co,Nantong 226010,China)
出处
《系统仿真学报》
CAS
CSCD
北大核心
2024年第8期1800-1809,共10页
Journal of System Simulation
基金
江苏省研究生科研与实践创新计划项目(SJCX23_2205)
2023江苏省工业和信息产业转型升级项目(苏财工贸〔2023〕60号)。
