Pothole detection using UAV with deep learning algorithm for road inspection

Pothole detection is a crucial component of road maintenance, essential for ensuring safety and minimizing vehicle damage. Traditional road inspection methods are often limited by their coverage, labor-intensive nature, and time-consuming processes. This paper presents an innovative approach to pothole detection by utilizing Unmanned Aerial Vehicles (UAVs) in combination with a Convolutional Neural Network (CNN) algorithm. The primary aim of this study is to evaluate the effectiveness of the CNN algorithm in detecting road potholes. The results indicate a high level of detection confidence, demonstrating that UAVs operating at low altitudes can accurately capture orthophotos for pothole identification. The pothole detection model achieved a precision of 0.437, a recall of 0.800, and a mean average precision (mAP) of 0.740, highlighting its accuracy and reliability. The study concludes that while UAVs integrated with artificial intelligence show promise for effective pothole detection, low-altitude flights present practical challenges due to environmental factors. Despite these limitations, the combination of UAVs and CNNs offers a viable solution for enhancing road inspection efficiency and accuracy.