| 朱达荣,齐永乐,汪方斌,王峰,左从菊,龚雪.融合距离自适应与密度感知的点云聚类算法[J].安徽建筑大学学报,2026,34(4):59-70 |
| 融合距离自适应与密度感知的点云聚类算法 |
| A Point Cloud Clustering Algorithm Integrating Distance Adaptation and Density Awareness |
| |
| DOI： |
| 中文关键词: 模式识别 激光雷达 点云聚类 DBSCAN算法 自适应机制 |
| 英文关键词: pattern recognition LiDAR point cloud clustering DBSCAN algorithm adaptive mechanism |
| 基金项目:安徽省自然科学基金面上项目（2508085MA020）；安徽省住建厅科学技术计划项目（2022-YF016、2022-YF065、2023-YF050）；安徽省教育厅研究生科研项目（YJS20210512）；安徽省新时代育人质量工程项目（2024xscx115）；机电类研究生军民融合联合培养基地项目（20231hpysfjd050） |
| 作者 | 单位 | | 朱达荣 | School of Electronic and Information Engineering,Anhui Jianzhu University,Hefei 230601,China；School of Mechanical and Electrical Engineering,Anhui Jianzhu University,Hefei 230601,China | | 齐永乐 | School of Electronic and Information Engineering,Anhui Jianzhu University,Hefei 230601,China | | 汪方斌 | School of Mechanical and Electrical Engineering,Anhui Jianzhu University,Hefei 230601,China | | 王峰 | Department of Information Engineering,Army Arms University of PLA,Hefei 230031,China | | 左从菊 | Department of Information Engineering,Army Arms University of PLA,Hefei 230031,China | | 龚雪 | School of Mechanical and Electrical Engineering,Anhui Jianzhu University,Hefei 230601,China |
|
| 摘要点击次数: 15 |
| 全文下载次数: 0 |
| 中文摘要: |
| 激光雷达点云密度随距离呈平方反比级衰减，造成点云空间分布不均匀。传统基于密度的空间聚类算法（Density-Based Spatial Clustering of Applications with Noise，DBSCAN）采用固定参数，在近场点云密集区域容易出现目标点云欠分割，在远场点云稀疏区域则容易出现目标点云过分割。为此，本文提出融合距离自适应与密度感知的DADS-DBSCAN算法（Distance-Adaptive Density-Sensitive DBSCAN）。该算法搭建密度感知阈值自适应模块与距离梯度自适应邻域模块，借助距离补偿因子与局部密度感知重构最小点数阈值和搜索半径，实现非均匀密度点云的自适应聚类；其次，引入主成分分析（Principal Component Analysis，PCA）几何约束驱动的簇合并模块，通过方向一致性校验合并过分割簇，修复目标几何完整性；最后，结合簇均值密度与空间孤立度搭建双重判定层次化精炼模块，精准滤除背景噪声。实验结果表明：该算法调整兰德系数（ARI）、归一化互信息（NMI）分别达到0.953 6、0.962 3，相较传统算法提升约7.06%、7.30%，戴维森堡丁指数（DBI）优化至0.483 5。实验证明，该算法能够有效解决近场欠分割、远场过分割问题，提升非均匀点云聚类精度与场景适配能力。 |
| 英文摘要: |
| The density of LiDAR point clouds exhibits an inverse-square decay with distance, leading to highly non-uniform spatial distribution. With fixed parameter settings, the traditional Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm suffers from under-segmentation of target point clouds in dense near-field regions and over-segmentation in sparse far-field regions. To address these problems, this paper proposes a Distance-Adaptive Density-Sensitive DBSCAN (DADS-DBSCAN) algorithm. The proposed algorithm constructs a density-sensitive threshold adaptation module and a distance-gradient adaptive neighborhood module. It adopts a distance compensation factor and local density perception to dynamically reconstruct the minimum point threshold and search radius, enabling adaptive clustering for non-uniform density point clouds. Furthermore, a geometric constraint-driven cluster merging module based on Principal Component Analysis (PCA) is introduced to merge over-segmented clusters through directional consistency verification and restore the geometric integrity of target objects. In addition, a dual-criteria hierarchical refinement module is established by integrating cluster mean density and spatial isolation to accurately remove background noise. Experimental results demonstrate that the Adjusted Rand Index (ARI) and Normalized Mutual Information (NMI) of the proposed algorithm reach 0.953 6 and 0.962 3, with improvements of approximately 7.06% and 7.30% compared with the traditional DBSCAN algorithm. Meanwhile, the Davies-Bouldin Index (DBI) is optimized to 0.483 5. The results verify that the proposed method effectively resolves the problems of near-field under-segmentation and far-field over-segmentation, and significantly improves the clustering accuracy and scene adaptability for non-uniform LiDAR point clouds. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
