Automating Post-processing of Terrestrial Laser Scanning Point Clouds for Road Feature Surveys

Terrestrial Laser Scanner (TLS) have seen increasing use in survey practices as a means of fast capturing large volumes of point data, with one such practice being the surveying of as-built features for roads. This paper presents methods to automate some of the processing of road scenes. The method focuses primarily on the automation of the identification and extraction process of kerbs and surrounding discrete geometric features, such as signage. The process is done in multiple stages. The first stage is to isolate the road surface and ground points using simple local classification and segmentation techniques. From this, it allows the locality of the points sampled from the kerb and road features to be approximated. For defining the kerb feature, the orientation of the kerb is found locally, and a 2D cross-section is extracted and examined to find the profile. The adjacent kerb profiles can then be joined to define a line representation along the kerb. For features adjacent to the road, points defined off the road surface are selected as candidate points. A region growing method is applied to group adjacent candidate points together to form group of points sampled from the same feature. This allows for the extraction of features such as signage, lights and poles from the scene. * Corresponding author.