Semantic Extraction of Roof Contour Lines From Airborne LiDAR Building Point Clouds Based on Multidirectional Equal-Width Banding

Building roof contours are important geometric features of the building and play an important role in the fields of three-dimensional model reconstruction and target detection. Due to the uneven density of the airborne light detection and ranging point clouds and the similar structure of the inner and outer contours of the building, the existing contour extraction algorithms have difficulty distinguishing between different contour structures. To solve this problem, a novel building contour line extraction method with semantic labeling is proposed in this article. First, the point clouds are divided into bands of equal width along multiple directions. All points in each band were projected onto the central axis of the band, the initial contour points were extracted according to the discontinuous distribution and farthest distance of projected points on the central axis. Second, a region composed of consecutive overlapping circles containing the initial outer contour points is constructed to identify the outer contour points within the region that have been classified as mixed contour points, and the remaining inner contour points are clustered and grouped using the density-based spatial clustering of applications with noise algorithm. Finally, each group of extracted contour points is subjected to sequentialization and densification to obtain the final contour lines. The proposed method was tested on the Tallinn, DublinCity, and Vaihingen datasets, and compared with other state-of-the-art algorithms. The results demonstrate that the proposed method has the advantages of being complete in structure, robust, and the contour extraction results are simple to express and contain semantic information.