Robust Pothole Detection in Adverse Weather Conditions using Thermal Imaging and Image Processing

Potholes on roads pose significant risks, contributing to road accidents and vehicle damage. Common strategies for detection, such as manual reporting and vibration-based sensors, have drawbacks like high setup costs and night vision limitations. This study explores the effectiveness of thermal imaging for pothole detection, utilizing a diverse dataset and applying augmentation techniques. The research exclusively employs a self-developed convolutional neural network (CNN) within deep learning. Results indicate a remarkable accuracy rate of 98%, surpassing pre-trained models. This study highlights the promising potential of thermal imaging in pothole detection, emphasizing its innovative application for improved road safety and maintenance.