THE USE OF MACHINE LEARNING MODEL FOR AUTOMATIC DETECTION OF ROAD TRAFFIC SIGNS

Subject Area: Computer Science

This study introduces an improved method for traffic sign detection and recognition, specifically designed for complex urban environments and varying lighting conditions. The proposed system is structured into three main phases: segmentation, shape classification, and recognition. In the segmentation phase, images are enhanced and segmented using the Hue, Saturation, and Value (HSV) color space, effectively isolating Regions of Interest (ROIs) despite the presence of similarly colored objects and lighting variations. The shape classification phase utilizes a Random Forest classifier combined with Distance to Borders (DtBs) features to accurately classify traffic signs into triangular, circular, and rectangular shapes. In the recognition phase, Random Forest and Support Vector Machine (SVM) classifiers are employed along with a range of feature extraction techniques, such as Histogram of Oriented Gradients (HOG), Gabor filters, Local Binary Patterns (LBP), and Local Self-Similarity (LSS). Experimental results on a publicly available dataset show high recall and accuracy, with an Area under the Curve (AUC) of 94.50%. These findings demonstrate that the proposed approach offers a reliable and efficient method for traffic sign detection and recognition, with potential applications in autonomous vehicles and real-time traffic monitoring systems. Future research will focus on integrating additional data sources and exploring advanced deep learning techniques to further enhance system performance.