Voxel-based representation of 3D point clouds: Methods, applications, and its potential use in the construction industry

Abstract

Point clouds acquired through laser scanning and stereo vision techniques have been applied in a wide range of applications, proving to be optimal sources for mapping 3D urban scenes. Point clouds provide 3D spatial coordinates of geometric surfaces, describing the real 3D world with both geometric information and attributes. However, unlike 2D images, raw point clouds are usually unstructured and contain no semantic, geometric, or topological information of objects. This lack of an adequate data structure is a bottleneck for the pre-processing or further application of raw point clouds. Thus, it is generally necessary to organize and structure the 3D discrete points into a higher-level representation, such as voxels. Using voxels to represent discrete points is a common and effective way to organize and structure 3D point clouds. Voxels, similar to pixels in an image, are abstracted 3D units with pre-defined volumes, positions, and attributes, which can be used to structurally represent discrete points in a topologically explicit and information-rich manner. Although methods and algorithms for point clouds in various fields have been frequently reported throughout the last decade, there have been very few reviews summarizing and discussing the voxel-based representation of 3D point clouds in urban scenarios. Therefore, this paper aims to conduct a thorough review of the state-of-the-art methods and applications of voxel-based point cloud representations from a collection of papers in the recent decade. In particular, we focus on the creation and utilization of voxels, as well as the strengths and weaknesses of various methods using voxels. Moreover, we also provide an analysis of the potential of using voxel-based representations in the construction industry. Finally, we provide recommendations on future research directions regarding the future tendency of the voxel-based point cloud representations and its improvements.