YOLOv8 and point cloud fusion for enhanced road pothole detection and quantification

Area of Science:

• Computer Vision
• Road Infrastructure Monitoring
• Intelligent Transportation Systems

Background:

• Automatic pothole detection is crucial for road maintenance and enhancing environmental perception in intelligent transportation systems.
• Reducing false positives is key to optimizing detection accuracy for road surface anomalies.
• Existing methods may struggle with irregular potholes and distinguishing them from other surface features.

Purpose of the Study:

• To introduce a novel method for detecting irregular potholes using integrated depth camera images and point cloud data.
• To improve the accuracy of pothole detection by effectively filtering out false positives.
• To provide precise measurements of pothole dimensions such as perimeter, surface area, and depth.

Main Methods:

• Utilizing YOLOv8 for initial 2D object detection to identify candidate pothole regions and associated 3D point clouds.
• Applying surface smoothness analysis to determine pothole boundary contours and extract relevant point cloud data.
• Implementing elevation thresholds to evaluate pothole depth and filter out non-pothole features like stains and patches.

Main Results:

• The proposed method demonstrated improved detection accuracy by [Formula: see text] on well-maintained roads compared to standalone YOLOv8.
• Achieved high performance metrics: precision of [Formula: see text], recall of [Formula: see text], and an F1 score of [Formula: see text].
• The model processes a single image in 0.23 seconds with low error rates for perimeter ([Formula: see text]), surface area ([Formula: see text]), and depth ([Formula: see text]).

Conclusions:

• The integrated approach of depth camera images and point cloud data offers a robust solution for accurate pothole detection.
• The method effectively reduces false positives, enhancing the reliability of automated road assessment.
• This technology contributes to more efficient road maintenance and improved intelligent transportation systems.