Artificial intelligence-powered smart roads: leveraging orange3 for traffic signs recognition

Arabiat, Areen; Altayeb, Muneera; Salama, Sanaa · 2025 · Crossref

DOI: 10.11591/ijai.v14.i5.pp3816-3826

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical need for accurate traffic sign recognition in advanced driver assistance systems (ADAS) and autonomous vehicles. While existing literature relies heavily on deep learning approaches like convolutional neural networks (CNNs), this research explores the efficacy of traditional machine learning classifiers implemented via the Orange3 data mining platform. The primary objective is to develop a robust model for detecting and classifying traffic signs to enhance road safety and traffic control efficiency. The experimental design utilized an open-source dataset from Kaggle containing 2,339 images of traffic signs categorized into eight classes: don't go, go, horn, roundabout, danger, crossing, speed limit, and unallowed signs. The methodology involved extracting 1,000 features from the images using an image embedder within Orange3. These features were fed into four supervised classification algorithms: Random Forest (RF), AdaBoost, Decision Tree (DT), and k-nearest neighbors (KNN). To prevent overfitting, the dataset was split into 70% for training and 30% for testing, with 10-fold cross-validation applied during model evaluation. Performance was assessed using metrics including accuracy, precision, sensitivity, specificity, F1-score, and the area under the receiver operating characteristic curve (AUC). The results demonstrated that ensemble methods significantly outperformed individual classifiers. Random Forest achieved the highest performance with an accuracy of 99.8%, an F1-score of 99.8%, and a specificity of 99.9%. AdaBoost followed closely with 99.2% accuracy and 99.9% specificity. Decision Tree and KNN yielded lower accuracies of 94.9% and 98.3%, respectively. ROC analysis confirmed the strong discriminative power of the models, with RF and KNN achieving perfect AUC values of 100.00%. When compared to recent studies utilizing CNNs and YOLO architectures, which reported accuracies ranging from 83.2% to 99.2%, the proposed RF model demonstrated superior classification accuracy. The study concludes that Random Forest is the most effective classifier for this specific traffic sign recognition task, offering high reliability and generalization capabilities. The findings highlight a positive correlation between model complexity and predictive accuracy, suggesting that ensemble methods are preferable for applications requiring high precision. The authors note limitations regarding environmental factors such as poor image clarity and background complexity, which can affect real-world performance. Future work is recommended to incorporate deep learning methods, multi-modal sensor data, and cross-domain adaptation techniques to improve robustness against varying conditions and hostile attacks, thereby enhancing the safety and operational efficacy of intelligent transportation systems.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• signage environment