标题：置信度引导的行道树提取并用于生态效益评估
Confidence guided roadside individual tree extraction for ecological benefit estimation
作者：Wen Fan, Bisheng Yang, Zhen Dong, Fuxun Liang, Jianhua Xiao, Fashuai Li
摘要：
行道树是城市生态系统的主要组成部分之一。从移动激光扫描（MLS）点云中提取路边的树木在城市生态学领域是一项热门研究，但由于严重的遮挡、街道设施和树桩系统的干扰以及树冠的重叠，这项工作仍然具有挑战性。为了克服这些限制，我们提出了一种置信度引导的行道树提取方法。首先，从道路场景中提取杆，并估计其可能是树干的置信度，来指导分割的顺序。其次，结合混合权重和异速生长模型约束的方法，优化的最小切割法被应用于单株行道树的提取。然后，在单树层面，树木形态学参数和三维绿量（LVV）被测量。最后，我们在四个具有不同点密度、树种、种植密度和遮挡的数据集上验证了所提出的方法，并得到了88.5%的精度和92.0%的召回率的结果，证明了所提出方法的稳健性和有效性，为街道场景提供了一种的生态评估思路。
本文的流程图（图1），置信度估算示意图（图二），分割结果（图3）如下所示；
图1. 流程图
图2. 置信度估计方法示意图（a）原始点云，（b）树干置信度估计
图3. 单树提取结果。I-IV代表数据集-IV。(a)代表整体情况。(b),(c),(d)是(a)的部分数字。
Abstract
Roadside trees are one of the main components of the urban ecosystem. The extraction of roadside trees from Mobile Laser Scanning (MLS) point cloud is essential in the field of urban ecology but is still challenging because of heavy occlusions, disturbance from tree stake systems, and overlapping of street furniture and crowns. To overcome these limitations, a confidence-guided roadside individual tree extraction method is proposed. Firstly, the pole-like objects are extracted from roadside objects and their confidence of being a trunk is estimated to guide the order of segmentation. Secondly, the optimized min-cut is applied for individual tree extraction by combining the confidence guidance and allometric growth model constraints. Thirdly, the morphological parameters and the living vegetation volume (LVV) are estimated at individual tree level. Finally, the proposed method was verified on four challenging datasets with different point densities, tree species, planting densities and occlusions, and achieves good performance in terms of 88.5% precision and 92.0% recall, demonstrating the robustness and effectiveness of the proposed method, and providing a meaningful tool for ecological assessment with the LVV of streets.
