Roof plane segmentation from airborne light detection and ranging (LiDAR) point clouds is an important technology for three-dimensional (3D) building model reconstruction. One of the key issues of plane segmentation is how to design powerful features that can exactly distinguish adjacent planar patches. The quality of point feature directly determines the accuracy of roof plane segmentation. Most of existing approaches use handcrafted features, such as point-to-plane distance, normal vector, etc., to extract roof planes. However, the abilities of these features are relatively low, especially in boundary areas. To solve this problem, we propose a boundary-aware point clustering approach in Euclidean and embedding spaces constructed by a multi-task deep network for roof plane segmentation. We design a three-branch multi-task network to predict semantic labels, point offsets and extract deep embedding features. In the first branch, we classify the input data as non-roof, boundary and plane points. In the second branch, we predict point offsets for shifting each point towards its respective instance center. In the third branch, we constrain that points of the same plane instance should have the similar embeddings. We aim to ensure that points of the same plane instance are close as much as possible in both Euclidean and embedding spaces. However, although deep network has strong feature representative ability, it is still hard to accurately distinguish points near the plane instance boundary. Therefore, we first robustly group plane points into many clusters in Euclidean and embedding spaces to find candidate planes. Then, we assign the rest boundary points to their closest clusters to generate the final complete roof planes. In this way, we can effectively reduce the influence of unreliable boundary points. In addition, to train the network and evaluate the performance of our approach, we prepare a synthetic dataset and two real datasets. The experiments conducted on synthetic and real datasets show that the proposed approach significantly outperforms the existing state-of-the-art approaches in both qualitative evaluation and quantitative metrics. To facilitate future research, we will make datasets and source code of our approach publicly available at https://github.com/Li-Li-Whu/DeepRoofPlane.

中文翻译：

---

用于屋顶平面分割的欧几里得和嵌入空间中的边界感知点聚类方法


来自机载光探测和测距 （LiDAR） 点云的屋顶平面分割是三维 （3D） 建筑模型重建的重要技术。平面分割的关键问题之一是如何设计能够准确区分相邻平面补丁的强大特征。点要素的质量直接决定了屋顶平面分割的精度。现有的大多数方法都使用手工制作的特征，例如点到平面距离、法线矢量等，来提取屋顶平面。但是，这些功能的能力相对较低，尤其是在边界区域中。为了解决这个问题，我们提出了一种在欧几里得和嵌入空间中的边界感知点聚类方法，由多任务深度网络构建用于屋顶平面分割。我们设计了一个三分支多任务网络来预测语义标签、点偏移量并提取深度嵌入特征。在第一个分支中，我们将输入数据分类为非屋顶点、边界点和平面点。在第二个分支中，我们预测将每个点移向其各自实例中心的点偏移量。在第三个分支中，我们约束同一平面实例的点应该具有类似的嵌入。我们的目标是确保同一平面实例的点在欧几里得空间和嵌入空间中尽可能接近。但是，尽管深度网络具有很强的特征表示能力，但仍然难以准确区分平面实例边界附近的点。因此，我们首先将平面点稳健地分组为欧几里得和嵌入空间中的许多集群，以找到候选平面。然后，我们将其余边界点分配给最近的集群，以生成最终的完整屋顶平面。 这样，我们可以有效地减少不可靠边界点的影响。此外，为了训练网络并评估我们方法的性能，我们准备了一个合成数据集和两个真实数据集。在合成和真实数据集上进行的实验表明，所提出的方法在定性评估和定量指标方面都明显优于现有的最先进的方法。为了促进未来的研究，我们将在 https://github.com/Li-Li-Whu/DeepRoofPlane 上公开提供我们方法的数据集和源代码。