Examining the Evolution of Structured Condition Index Using Linear Regression and Automated Classification of Inspection Results by CNN

¹ Applied Geosciences Laboratory, Faculty of Science, Mohammed First University, Oujda, Morocco
² Modeling and Combinatorics Laboratory, Polydisciplinary Faculty Safi (PFS), Cadi Ayyad University, Safi, Morocco
³ Mathematics, Signal and Image Processing, and Computing Research Laboratory (MATSI), Ecole Supérieure de Technologie Oujda (ESTO), Mohammed First University, Oujda, Morocco
⁴ General Roads Directorate, Ministry of Equipment and Water, Rabat 10100, Morocco

^* Corresponding author: youssef.aouni@ump.ac.ma

Abstract

Monitoring pavement deterioration is a major concern for road network organizations worldwide. If left unaddressed, deterioration can impact safety and incur additional costs. This is why continuous monitoring of pavement condition is one of the most important management methods. Deterioration is reflected in various performance indices, including the Structured Condition Index (SCI), which is used in Morocco. This study aims to analyses how the condition of the structure layer has changed and to predict possible variations in the surface indicator on Road 16 in the Oriental Region between 2010 and 2024. The data comprises images of pavement deterioration collected from the General Directorate of Roads' archive in Morocco. These include records of deflection, pavement uniformity and the distribution of the three most prevalent types of pavement deterioration. The images of pavement deterioration were automatically classified using the DenseNet121 architecture with 92% accuracy. A second classification of the images was then performed using DenseNet201 and a grid ranging from A to D to quantify the severity of the degradation. An analysis of the SCI variation curves was conducted, followed by regression-based prediction. This research continues to support road managers in their decision-making processes.