David Belton

Classification and Segmentation of 3D Terrestrial Laser Scanner Point Clouds

DavidBelton
University
Curtin University
Supervisor (Academic)
Dr Derek Lichti, Curtin University (now at Calgary)
Supervisor (Industry)
Chris Earls, AAM
Projects
mysite
Employment
Research Fellow, Curtin University
Thesis Abstract

With the use of terrestrial laser scanning, it is possible to efficiently capture a scene as a 3D point cloud. As such, it is seeing increasing deployment in traditional surveying and photogrammetric fields, as well as being adapted to applications not traditional associated with surveying and photogrammetry. The problem with utilising the technology is that, since the point cloud captured is so densely populated, the processing of the data can be extremely labour-intensive. This is due to the large volume of data that must be examined to identify the features sampled and to remove extraneous information. Research into automated processing techniques aims to alleviate this bottleneck in the work-flow of terrestrial laser scanner (TLS) processing.

A segmentation method is proposed in this thesis to identify and isolate the salient surface that comprises a scene sampled as a 3D point cloud. The cut-plane based region growing (CPRG) segmentation method uses the classification results, approximated surface normals, and the directions of principal curvature to locally define the extents of the surfaces present in a point cloud. These generalised surfaces can be of arbitrary structure, as long as they satisfy the imposed surface conditions. These conditions are that, within the identified extents of the surface, the surface is considered to be continuous and without discontinuities. As such, a novel metric is introduced to determine points sampled near discontinuous or changes in the surface structure that is independent of the underlying structure of the surfaces. In addition, an iterative method of neighbourhood correction is also introduced to remove the effects of multiple surfaces and outliers on the attributes calculated through the use of local neighbourhoods. 

The CPRG segmentation are tested on practical 3D point clouds captured by a TLS. These point clouds contain a variety of different scenes and objects, as well as different resolutions, sampling densities, and attributes. It was shown that the majority of surfaces contained within the point clouds are isolated as long as they have a sufficient sampling to be resolved. In addition, different surfaces types, such as corrugated surface, cylinders, planes and other complex smooth surfaces, are segmented and treated similarly, regardless of the underlying structure. This illustrates the CPRG segmentation method’s ability to segment arbitrary surface types without a prior knowledge.