Automatic detection and classification of road defects on a global-scale: Embedded system

Road networks are created with different pavement designs. In general, flexible pavement is preferred as the most used superstructure type in the world. Road networks with this superstructure experience deterioration for some reasons. These deteriorations take different forms over time and create road defects. The process of identifying defects is very important for the efficiency of the road pavement and traffic components. In this study, the process of automatic detection and classification of road defects occurring in flexible superstructure road networks was carried out with an intelligent system. Road images obtained from eight different countries and 10 different road defects were taken into consideration in the study. YOLOv5 object detection model was used in the detection process of defects. The training processes were completed by creating five different variants and different combinations and YOLOv5l provided the best performance. mAP and F1 Score values were determined as 0.526 and 0.540, respectively. In addition, a global-scale automatic road defect detection system has been developed via new data set consisting of different countries. The developed system is a prototype and has the ability to detect and classify defects occurring in road networks in real-time. To the best of our knowledge, this is the first study to detect the most road defect classes in real time. The created system was tested on a 4.5 km long university campus network as a case study. D00-517, D10-507, D20-50, D30-2, D40-48, D50-18, D60-343, D7025 and D80-9 were detected in real-time by the embedded system. It is clear that the system will be a guide for road network managers by obtaining location information of the identified defects. Detecting, classifying and locating road defects with the developed system will accelerate the maintenance and repair process of road networks and also extend their service life. Road safety and comfort of traffic components using the road network will also be increased. As a result, an example of vehicle-infrastructure (V2I) communication, which is a form of intelligent transportation systems application, is presented in this article.