Heuristic Approach to Identifying Horizontal Curves and Their Parameters Given Lidar Point Cloud Data

Abstract

Information on the location and characteristics of highways has many applications to highway engineering, such as highway safety analysis, asset management, and highway improvement planning. However, the task of creating an inventory of highway curves in a state is a daunting task, which cannot be done efficiently by field visits. The alternative approach that processes light detection and ranging or GPS data sets to identify highway curves has shown promise as a more efficient method but has also proved to be a difficult task. This paper presents a procedure for curve detection and analysis and also examines the success rate for identification and classification of those curves identified by this method. The basic process began with the roadway divided into segments, each with assigned attributes. The segments with similar attributes were combined into larger sections, each classified as either a curve or a tangent. Segments that were too short to be a true curve were removed. The heuristic analysis used to optimize threshold values for the analysis of these curves was also discussed. It was found that the curve identification success rate of this process, depending on calibration, ranged from 84.4% to 92.9% accuracy (e.g., the segment identified correctly as a curve or tangent), and curve geometries (e.g., points of curvature or points of tangency location correctly found) could be expected to have an accuracy rate of 78.7% to 89.9%.