Pothole Detection and Dimension Estimation via Image Transformation and Scaling with Thai Road Data Integration

Abstract. Potholes on roads affect traffic safety and the overall quality of infrastructure. If left unrepaired, they can lead to increased maintenance costs and broader community impacts. Traditional inspection methods, such as visual surveys by human observers, still have limitations in terms of efficiency, accuracy, and safety. To facilitate manual inspection process, a pothole detection and dimension estimation technique combining deep learning and image processing techniques is presented in this study. The method employs the YOLOv8n-seg model, which performs instance segmentation to outline pothole boundaries. Model training was conducted using a combination of open-source and Thai roads pothole dataset to enhance contextual relevance. Inverse Perspective Mapping (IPM) was applied to estimate pothole dimensions and convert front-view images into bird's-eye views. The segmentation masks predicted by the model were then used to calculate the real size of each pothole. The presented method requires single camera calibration for each camera installation. The results highlight the potential of integrating deep learning with image processing techniques to support road condition monitoring, as evidenced by a precision of 0.745, mAP@0.5 of 0.708 and an average dimension estimation error of 11.30%.