Traffic Light State Recognition Based on YOLO Detection and Circle-Based Structural Analysis

Traffic light recognition is a key element in determining vehicle movement, playing a very important role in autonomous driving systems. This paper proposes a hybrid traffic light recognition method that combines deep learning-based object detection and circle-based color classification. First, the YOLO model is used to accurately find the location of traffic lights in various environments. The detected area is set as a Region of Interest (ROI), and then the circular traffic light structure is identified through grayscale conversion and Hough Circle Transform. Afterward, each circular region is separated through thresholding and masking processes. Based on this, the illuminated lamp is precisely re-detected, its position is compared, and its state is determined. Unlike existing HSV or RGB threshold-based methods, the proposed technique has the advantage of being robust to changes in illumination or occlusions. Experimental results show that this method stably recognizes red, yellow, and green signals in real-world environments. The approach in this paper is expected to contribute to helping autonomous vehicles comply with traffic laws and enhancing road safety.