Recognition of Sri Lankan traffic signs using machine learning techniques

Abstract

The recognition of traffic signs is a crucial component of driver assistance systems that have been extensively researched worldwide. However, it remains a challenging issue due to the increasing number of vehicles, road signs, and the lack of awareness among drivers and other road users. A Traffic Sign Recognition (TSR) system is an advanced autonomous technology designed to assist drivers by accurately identifying and interpreting traffic signs. This system plays a crucial role in enhancing driver awareness and ensuring appropriate responses to various traffic conditions. The precise recognition of traffic signs is essential for maintaining road safety and improving the overall driving experience. This study focuses on the recognition of Sri Lankan traffic signs and examines the combination of classifiers with a specific feature extractor. A dataset of 300 images of road signs was utilized for this study by capturing the images. The Scale Invariant Feature Transform (SIFT) was used as a feature descriptor in this process. The classifiers employed were Support Vector Machine (SVM) and k-Nearest Neighbor (k-NN). Different combinations of SVM and k-NN were applied to the dataset, and the study achieved 100% accuracy with various combinations of k-NN. The study found that the combination of SIFT and SVM is the most effective method for the proposed recognition of traffic signs.