3D profile-based pothole segmentation and quantification

With the increasing traffic load, pavement distresses are caused inevitably. Water penetration and extreme weather condition speed up the deterioration of pavements and cause the occurrence of potholes. Automated pothole inspection methods have been developed with both 2D and 3D-based imaging techniques for many years. However, the performances suffer from either accuracy or efficiency. In this paper, a 3D profile-based solution is proposed to inspect potholes with high accuracy and efficiency. A low-cost stereo imaging system is deployed to generate the 3D pothole profile, and an algorithm integrating region growing is developed to segment potholes. The pothole volume is calculated based on the segmentation results and the depth information. Overall, the proposed method outperforms the existing method by 1.72% and 5.192% in pothole segmentation and quantification, respectively. Moreover, the proposed method has no demand for large-scale datasets and training procedures, thus reducing time and labour costs.