Automatic extraction of pole-like objects from different scene point clouds
Wen Hao, Tianwang Luo, Wei Liang · 2025
Pole-like objects (PLOs) play an important role in road safety and planning. Accurate extraction of PLOs is essential for effectively preventing dangerous situations during driving. In this paper, we propose an automatic PLO extraction method based on geometric features and proximity relations. First, PLOs are accurately located by leveraging their linear distribution and approximately circular cross-section after geometric feature calculation and ground filtering. Then, a region-growing method based on 17 neighborhoods is proposed to segment the overlapping components of adjacent PLOs using the distance information in 2D space. Subsequently, a voxel distribution analysis-based method is proposed to accurately segment the overlapping parts. Finally, characteristics such as area variation and height are applied to refine the segmentation results. Experiments are conducted on different scenarios to evaluate the effectiveness of the proposed method. According to the experimental results, our proposed method can achieve good extraction performance in both urban street and forest scenes.