LiDAR-Based Assessment of Roadside Slope Hazards: Enhancing Safety on Rural Roads (2024)

Undergraduate: Saurya Acharya

Faculty Advisor: Ashok Krishnamurthy
Department: Computer Science

---

A significant proportion of motor vehicle fatalities occur across the United States, especially among road users in rural areas. This study focuses on identifying potentially hazardous roadside slopes along secondary roads in rural regions, aiming to mitigate risks and enhance transportation infrastructure safety. Utilizing existing topographical survey data from the North Carolina Department of Transportation (NCDOT), aerial Light Detection and Ranging (LiDAR) scans were used to develop a methodology for analyzing roadside slopes. Three-dimensional LiDAR point clouds were rasterized into two-dimensional, top-down images of a roadway scene. The image was processed using variable reconstruction techniques and the roadway was segmented by multiple edge detection methods. Then, slope-fitting was tested for LiDAR points adjacent to road segments, using the R-squared value as a metric, with variable length and width to determine the optimal dimensions. The optimal segment found was 3 edge pixels (2.4 - 3.4 ft) long with lateral measurement areas extending 7 pixels (5.6 ft) from either road edge. We were able to identify roadside slopes adjacent to 92.0588% of roadway edge pixels.