AI-Driven Automated Pothole Detection and Geospatial Mapping System for Real-Time Road Condition Monitoring

Abstract

Road surface degradation, particularly potholes, poses significant risks to road safety, vehicle longevity, and infrastructure planning. This paper presents an AI-based automated pothole detection and geospatial mapping system designed to monitor real-time road conditions. The system operates by detecting potholes as the vehicle equipped with the device passes by, capturing a snapshot of the pothole in the frame, and recording the timestamp and GPS coordinates. The data is then stored in a database, providing a detailed record of pothole locations. This information can be leveraged by government agencies for more efficient and cost-effective infrastructure maintenance, as it enables quicker identification and repair of road defects. The system employs the YOLO26 object detection model, selected for its computational efficiency and suitability for deployment on resource-constrained edge devices. To improve reliability in real-world scenarios, the system incorporates geospatial tagging, duplicate detection elimination through spatial–visual clustering, and probabilistic confidence aggregation to reduce false detections. The integrated architecture further supports interactive visualization through a web-based dashboard for monitoring detected potholes and managing repair status. YOLO26 was benchmarked against YOLOv8s with identical datasets, training configurations, and hyperparameters. Results show YOLO26 achieving a mean Average Precision (mAP@0.5) of 78.9%, surpassing YOLOv8s (76.96%) with higher precision (85.0% vs. 81.0%) while maintaining similar recall. The system also integrates GPS-tagged geospatial mapping, enabling probabilistic monitoring where multiple vehicle observations enhance coverage and reliability. This study is among the first to apply YOLO26 for road condition monitoring, offering a scalable and efficient solution for large-scale road condition monitoring and infrastructure maintenance planning.