Abstract
Raster grid, triangulated irregular network (TIN), point cloud, and octree are the commonly used methods to represent scattered light detection and ranging (LIDAR) data for building detection. Compared to these representations, the benefit of voxel representation lies in that the implicit notion of adjacency and the true 3D representation can be presented simultaneously. Based on the binary voxel representation of LIDAR data, a voxel-based 3D building detection (V3BD) algorithm is proposed to separate building and non-building voxels. The proposed V3BD algorithm consists of three steps: (1) regularizing LIDAR point clouds into binary 3D voxel structure (B3VS); (2) removing ground voxels using a voxel-based 3D filtering algorithm; and (3) selecting a group of non-ground voxels with straight-line and elevation jump characteristics as seeds and then labeling them and their 3D connected sets as building voxels. ISPRS urban datasets, which are representative of buildings of diverse types, are used to analyze the sensitivity of “adjacency size” parameter in the B3VS model and assess the accuracy of V3BD algorithm quantitatively. The quantitative evaluation results indicate that: (1) The 56-adjacency is the optimal adjacent size; (2) for completeness and correctness indexes, average values of 96.11% and 95.87% for buildings are obtained, respectively. The qualitative evaluation results indicate that large, dense, and irregularly shaped buildings or buildings with eccentric roofs are all successfully captured. The proposed algorithm is promising for building detection in automatic fashion. The detected building results can directly serve as building model that is a new form of 3D voxel model with certain accuracy.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call