Roof Plane Segmentation From LiDAR Point Cloud Data Using Region Expansion Based L ₀ Gradient Minimization and Graph Cut

Abstract

Automatic roof segmentation from airborne light detection and ranging (LiDAR) point cloud data is a key technology for building reconstruction and digital city modeling. In this paper, we develop a novel region expansion based L0 gradient minimization algorithm for processing unordered point cloud data, and a two-stage global optimization method consisting of the L0 gradient minimization and graph cut for roof plane segmentation. Firstly, we extract the LiDAR points of buildings from the original point cloud data with a deep learning based method and separate the points of the different buildings using Euclidean clustering to improve the processing efficiency. Secondly, region expansion based L0 gradient minimization is proposed, which is specially designed for roof plane segmentation from unordered point clouds. To fundamentally avoid the need for empirical parameter tuning in L0 gradient minimization, we propose a multi-stage coarseto-fine segmentation process, which further improves the effect of the roof plane segmentation. Finally, graph cut is utilized to solve the jagged boundary and over-segmentation problems existing in the segmented roof planes and produce the segmentation results. We conducted comparative experiments on the Vaihingen and Hangzhou datasets. The experimental results show that the proposed approach significantly outperforms the current state-of-the-art approaches at least 6.7% and 8.9% in roof plane quality index in the Vaihingen and Hangzhou datasets, while showing superior robustness to different kinds of data.

GeoAI: a review of artificial intelligence approaches for the interpretation of complex geomatics data

“…Moreover, to demonstrate the accuracy of their method, they present a performance comparison with other state-ofthe-art methods. Papers like Zhang and Zhang (2017), Wang and Ji (2021), and Lee et al (2021) make extensive use of approaches based on DL for semantic parsing of 3D point clouds of urban building scenes.…”

Combining Deep Semantic Edge and Object Segmentation for Large-Scale Roof-Part Polygon Extraction from Ultrahigh-Resolution Aerial Imagery

“…Undeniably, deep learning has become the tool of choice for the automated supervised image analysis of airborne and spaceborne Earth observation (EO) imagery, be it for object detection, semantic segmentation, or instance segmentation [2]. Largely driven by data availability, the majority of research concerned with applying deep learning for rooftop mapping focuses on either (1) building footprint extraction from RGB (ortho-)images [3], sometimes combined with LiDAR-derived 2D height maps [4,5], or (2) roofplane segmentation from LiDAR point clouds [6][7][8][9][10]. Research on the first application is predominantly propelled by well-known open-source datasets such as Vaihingen [11], the Inria Aerial Image Labelling Dataset [12] or the more recent SemCity Toulouse [13].…”