Manuscript received April 3, 2025; revised April 16, 2025; accepted June 3, 2025; published August 7, 2025.

Abstract—Road Surface Conditions (RSC) play a critical role in daily transportation and infrastructure reliability. Due to heavy public, governmental, and industrial dependence on road networks, maintaining road quality is essential. However, varying weather conditions often lead to road damage, complicating the task of accurately identifying RSC. To address this challenge, we propose a hybrid model combining Deep Learning (DL), Optimization, and Machine Learning (ML) techniques for effective recognition and classification of RSC. The model is developed using a publicly available dataset of road surface images. Preprocessing is performed using Wavelet Transform (W), followed by the extraction of texture features using Gray Level Co-occurrence Matrix (G). Deep feature extraction is conducted using the Efficient-Net (E) model. The resulting features are then optimized using the Firefly Optimization (F) algorithm. Finally, classification is carried out using a Support Vector Machine (SVM). This approach enables accurate RSC identification with a minimal number of data points while maintaining high performance. The combined DL-ML framework demonstrates superior results in terms of key evaluation metrics such as sensitivity, specificity, precision, and accuracy. The proposed model achieves an accuracy rate of 99.38%, specificity 99.3%, sensitivity 99.4% and precision 99.36% respectively. The PBIAS obtained using proposed model is 0.0687%.

Keywords—road surface condition, Wavelet Transform (W), Efficient-Net, firefly optimization, Support Vector Machine (SVM)

Cite: Ramya K. Rajavolu and Nutakki Jyothi, "WGEF: An Optimized Deep Learning Model for Recognition of Road Surface Condition Using SVM Classifier," Journal of Image and Graphics, Vol. 13, No. 4, pp. 427-436, 2025.

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.