Segmentation of tree seedling point clouds into elementary units

This article describes a new semi-automatic method to cluster terrestrial laser scanning (TLS) data into meaningful sets of points to extract plant components. The approach is designed for small plants with distinguishable branches and leaves, such as tree seedlings. It first creates a graph by connecting each point to its most relevant neighbours, then embeds the graph into a spectral space, and finally segments the embedding into clusters of points. The process can then be iterated on each cluster separately. The main idea underlying the approach is that the spectral embedding of the graph aligns the points along the shape’s principal directions. A quantitative evaluation of the segmentation accuracy, as well as of leaf area (LA) estimates, is provided on a poplar seedling mock-up. It shows that the segmentation is robust with false-positive and false-negative rates of around 1%. Qualitative results on four contrasting plant species with three different scan resolution levels each are also shown.

本文提出一种全新的半自动方法，可将地面激光扫描（terrestrial laser scanning, TLS）点云数据聚类为具有语义关联的点集，以提取植物各组分。该方法专为具备可区分枝条与叶片的小型植物设计，例如树木幼苗。其流程首先通过将每个点与其最相关的邻点相连构建图结构，随后将该图嵌入至谱空间，最终将嵌入结果分割为若干点簇。后续可针对每个点簇单独迭代执行上述流程。该方法的核心思想在于，图的谱嵌入可将点云沿目标物体形状的主方向进行对齐。本文针对杨树幼苗模型开展了分割精度与叶面积（Leaf Area, LA）估算结果的定量评估，结果显示该分割方法鲁棒性优异，假阳性率与假阴性率均约为1%。此外，本文还展示了针对4种不同植物种类、每种植物分别采用3种不同扫描分辨率所得到的定性分割结果。