Related Experiment Videos
Smartphone-based accurate pothole depth estimation using monocular RGB and LiDAR-guided deep learning.
Waqar Rauf Butt1, Muhammad Farooq1, Sohail Jabbar2
1Department of Information Technology, University of the Punjab, Lahore, Pakistan.
Plos One
|May 8, 2026
Summary
This study introduces a low-cost smartphone framework using LiDAR for accurate pothole detection. It enables efficient, real-time road condition assessment for infrastructure management.
Area of Science:
- Civil Engineering
- Computer Science
- Geospatial Technology
Background:
- Road infrastructure maintenance is crucial for safety and transportation efficiency.
- Potholes pose a significant challenge to continuous road monitoring.
- Existing detection methods are often costly and impractical for large-scale, real-time use.
Purpose of the Study:
- To develop a low-cost, scalable framework for pothole assessment.
- To leverage smartphone LiDAR sensors for automated road condition monitoring.
- To improve the accuracy and efficiency of pothole detection and delineation.
Main Methods:
- Utilized a hybrid deep learning pipeline fusing RGB imagery with LiDAR data.
- Implemented LiDAR-guided refinement for precise depth estimation and segmentation.
- Developed a post-processing step for enhanced boundary alignment.
- Curated a dataset of 25,000 RGB-LiDAR image pairs for training and validation.
Main Results:
- Achieved high accuracy in depth estimation and pothole segmentation.
- Demonstrated superior performance compared to existing pothole detection methods.
- Validated the framework's effectiveness across diverse road and lighting conditions.
Conclusions:
- Smartphone-based LiDAR offers a practical and economical solution for real-time road assessment.
- The proposed framework provides a valuable tool for modern infrastructure management.
- Enables efficient and automated monitoring of road conditions to enhance public safety.
