Hybrid Model for Realistic and Efficient Estimation of Highway Sight Distance Using Airborne LiDAR Data
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
2019 American Society of Civil Engineers. High-precision light detection and ranging (LiDAR) data, which provide a rather close fit to the real three-dimensional (3D) environment, have been used to conduct a computerized estimation of sight distance along existing highways in place of risky, cumbersome field measurements. However, so far, there is no method for estimating highway sight distance both efficiently and realistically based on high-density LiDAR data. To tackle this problem, a hybrid model supported by MATLAB version R2018b, which combines an accurate algorithm of visibility analysis for the modified Delaunay triangulation (MDT) method with a back-propagation (BP) neural network, was developed to analyze highway sight distance using airborne LiDAR data. The MDT algorithm not only enables an accurate determination of the visibility of any given object point, but also allows for the collection of numerous labeled training data for the neural network. Based on a training set with 416,721 data points, a neural network was constructed and incorporated into the model to improve the efficiency of sight distance computation. The results demonstrate that the hybrid model serves as an effective tool for highway sight distance evaluation using LiDAR data. Compared with existing methods, a more realistic analysis of highway sight distance is provided by the hybrid model. Also, the computational efficiency of estimating sight distance is improved to a satisfactory level, which is beneficial to practical applications in large-scale, real-world projects.
