Automated Road Crack Detection Using Deep Convolutional Neural Networks

Effective and timely identification of cracks on the roads are crucial to propitiously repair and limit any further degradation. Till date, most crack detection methods follow a manual inspection approach as opposed to automatic image-based detection, making the overall procedure expensive and time-consuming. In this study, we propose an automated pavement distress analysis system based on the YOLO v2 deep learning framework. The system is trained using 7,240 images acquired from mobile cameras and tested on 1,813 road images. The detection and classification accuracy of the proposed distress analyzer is measured using the average F1 score obtained from the precision and recall values. Successful application of this study can help identify road anomalies in need of urgent repair, thereby facilitating a much better civil infrastructure monitoring system. The codes associated with the study including the trained model can be found in 11.