Intelligent Road Condition Monitoring and Analysis for Safer Highways

Road damages are one of the greatest threats to the safety of road users as they contribute to mechanical breakdowns and eventual accidents. This project addresses the problem by presenting a possible solution of how to develop an efficient road damage detection system using the YOLOv5 object detection algorithm. The system makes use of the images from the road damage datasets to create a YOLOv5 model that is made to detect potholes, joint cracks, and alligator cracks and outlines where the potholes are with bounding boxes. Furthermore, GPS coordinates of the detected damage are logged and stored in cloud database and this can be used to provide road management authorities information on hazardous road conditions for action. The system also investigates pothole depth prediction through MiDaS, an emerging method in the area of Depth Estimation. The proposed system attains an average of 78.1% detection accuracy and a mAP@50 of 76.6% on average. These results validate the system’s real-time pothole and crack detection and classification performance while setting the foundation for enhanced detection sensitivity, depth prediction, and navigation integration for real-time pothole warning to drivers. By integrating geo-tagging based on IoT, the effectiveness of performing anticipatory repair and maintenance of road infrastructure is highly increased, therefore road safety is improved.