Rural Road Pavement Panel Detection with Unmanned Aerial Vehicles

Since they connect major population and economic hubs across borders, roads and transportation play a crucial role in a nation's economy. Crash and accident rates are a direct result of poor and substandard road conditions. Road inspectors can better plan repair and assessment tasks with the help of the Pavement Maintenance System (PMS). Distress detection surveys are still carried out manually in developing nations. However, manual road inspection techniques carried out on the ground are dangerous, time-consuming, and labor-intensive. Unmanned aerial vehicles (UAVs) offer the flexibility and manoeuvrability to swiftly visually inspect the damaged area and gather the required data without endangering human safety. Surveyors would therefore be better able to assess the state of road pavement in rural areas if pavement panel detection using unmanned aerial vehicles (UAVs) was implemented. In this study, the precision, recall, and latency of the YOLOv7 and YOLOv8 instance segmentation models for pavement panel detectors were examined. This paper also introduced a new dataset of UAV images from complex and unstructured road environments within the highway areas in the Philippines. The findings indicate that YOLOv8 performed better, with a latency of 84.65 ms, a precision of 97.36% ±0.002, and a recall of 98.71 %±0.004. Road users and managers may benefit from an accelerated pavement assessment process in rural areas thanks to the suggested pavement panel detection model.