Enhanced Road Surface Classification using CNN and Random Forest Models

This study presents the development and evaluation of a road surface categorization model designed to classify surfaces into four categories: dry, wet, snowy, and muddy roads. The model, trained on a dataset of 11,900 images, achieved an overall accuracy of 92.43 %. For dry roads, the proposed model demonstrated a precision of 92.59% and a recall of 90.91 %, indicating robust performance in identifying and correctly classifying dry road surfaces. Similarly, the classification of wet, snowy, and muddy roads showed strong results, with precision and recall scores ranging from 91.23% to 93.75% and 90.91 % to 93.75%, respectively. The proposed model's overall accuracy of 96% underscores its efficacy in categorizing diverse road conditions. Macro, weighted, and micro average scores align closely with individual class performances, reinforcing the model's reliability and consistency. These findings suggest significant implications for infrastructure management, road safety, and transportation efficiency, showcasing the model's potential applications in intelligent transportation systems, route planning, and road maintenance strategies through advanced machine learning techniques.