Vision- and Entropy-Based Detection of Distressed Areas for Integrated Pavement Condition Assessment

Abstract

Pavement management systems aim to secure roadways functionality and vehicle passengers’ safety by proposing strategies for pavement assessment and maintenance. However, transportation departments lack accurate, low-cost, and efficient methods for pavement assessment. Presented in this paper is a vision-based system for the detection of distressed pavement areas using low-cost technologies. Videos of pavement surface are recorded by a camera placed at the rear of a passenger vehicle, moving in a real-life urban network under normal traffic conditions. Collected data is processed by a developed algorithm that identifies video frames, including any type of pavement defect, using image entropy with a frame-based classification accuracy, precision, recall, and F1 score of 89.2%, 86.6%, 85.6%, and 86.1%, respectively. The proposed system can serve as the basis of any integrated pavement management system, saving significant amounts of time and cost for transportation departments.