U-Net-based integrated framework for pavement crack detection and zone-based scoring
Mohammadreza Sabouri, Alireza Sepidbar
Abstract
Pavement scoring systems are cost-effective means of integrating the pavement management to prevent the progression of failure. The current research developed a pavement scoring system in which the presence and severity of longitudinal and transverse cracks in asphalt pavements are detected through digital images. The pavement cracks were identified using U-Net-based network architecture optimised by the grasshopper optimisation algorithm. The quantum geographic information system was used to reveal the cracks on the study zones. These zones then were evaluated, compared and scored according to their mean crack length, mean crack width and the number of cracks. The analytic hierarchy process based on expert opinion was employed to determine the importance and weight of these factors. Each area received a score based on the condition of the pavement that enabled comparison and identification of the critical areas. The proposed method can allow transportation agencies to identify and evaluate critical areas and determine the proper maintenance and repair priorities.