Automatic Framework for Detecting Obstacles Restricting 3D Highway Sight Distance Using Mobile Laser Scanning Data

Abstract

Periodic measurements of sight distance on as-built roads and subsequent removal of sight obstructions are important for guaranteeing highway safety. In this paper, an accurate and efficient framework is proposed for automated detection of obstacles restricting sight distance on highways using mobile laser scanning (MLS) data. The developed framework was implemented in MATLAB (version: 2020a) and operates along the mapping trajectory recorded in the MLS data. A linear index-based segmentation technique was used to efficiently segment MLS point clouds; based on this, methods for identifying target points, removing on-road noise, and detecting sight obstacles were then developed. The target points for sight obstacle detection were derived from the pavement surface points, which were identified via a similarity-and-connectivity-based technique. Considering that on-road vehicle noise may adversely affect the detection of sight obstructions, a data-refinement procedure was developed to remove them and to fill the missing point regions they caused. For each sight point in the mapping trajectory, a segmentation-based algorithm was applied to achieve fast sight obstacle detection. Tests on MLS data from two real-world highways in the case study showed that the proposed framework detected sight obstructions on the combined highway alignments in the presence of noise. The procedure detected sight obstacles at each sight point within 0.2 s, with limited computational power. Therefore, it can be applied in real-world projects and will be of interest to researchers and practitioners in this field.