Comparison of Methods Used for Detecting Unknown Structural Elements in Three-dimensional Point Clouds

Three-dimensional (3D) imaging technologies, in particular 3D laser scanners, are becoming more accessible and more accurate. These advances are providing engineers and architects with vast quantities of raw, geometric data. Whereas this data is visually appealing and intuitive to the human eye, it contains very little meaning beyond that. The research presented in this paper presents and compares methods for attributing meaning to dense 3D point clouds. Two of the methods developed and presented utilize 2D and 3D Hough transforms to represent the points as lines and planes. The third method uses point segmentation techniques to group points belonging to the same plane. The initial focus is on structural steel systems and connections modeling for analysis of reuse. The advantages and disadvantages of each method are outlined, and each method is evaluated for its potential to provide engineers and architects with useful and meaningful point clouds from 3D laser scanners. The point segmentation techniques exhibit the most potential by allowing for the location and orientation of any surface to be identified.