Paper Details

Name: Sparsh Verma, Shashank Kumar, Ayush Jindal & Sandeep Kumar
Country: India

Year: 2026
Volume: Volume No: 13, January, Year: 2026 (Special Issue)
Page Number: 471-480
DOI: https://doi.org/10.5281/zenodo.19063123

ABSTRACT
This study presents an automated approach for detecting potholes and estimating their depth using LiDAR point cloud data. The methodology involves pre-processing raw LiDAR data, segmenting the road surface, clustering pothole regions, and calculating their
dimensions. The proposed system utilizes DBSCAN clustering and convex hull techniques to accurately identify and measure potholes. The methodology involves collecting 3D point cloud data from LiDAR sensors, specifically from .las files, to ensure accurate surface reconstruction. The pre-processing stage applies voxel down sampling, statistical noise removal, and plane segmentation using RANSAC to differentiate road surfaces from potholes. Henceforth, an algorithm called DBSCAN is then employed that identifies the regions in which the potholes are present. Then an unambiguous extraction of pothole dimensions is done using the convex hull estimation. This methodology not only provides an accurate measurement of the pothole depth but also aids in real-time view of the potholes, which is important for the maintenance of roads and ensuring road safety.

General Terms
Algorithms, Measurement, Computer Vision, Data Processing, Remote Sensing

Keywords
Computer Vision, DBSCAN clustering, LiDAR, Point Cloud, Potholes, RANSAC