基于YOLO的交通标志识别研究

Research on Traffic Sign Recognition Based on YOLO

摘要: 在交通标志识别过程中，经常受到复杂多变的道路环境、动态光照条件、交通标志的多样化设计以及实时性要求的影响，导致识别效果不够理想。本文基于YOLOv8算法，采用TT100K数据集进行交通标志识别训练和测试。结果表明，YOLOv8算法在交通标志识别中展现出较好的性能，对小目标交通标志牌、复杂背景中的交通标志牌、亮度分布不均匀背景中的交通标志牌等也能够取得比较满意的识别效果，为后续交通标志识别算法的优化及多场景应用研究奠定了技术验证基础。

Abstract: In the process of traffic sign recognition, recognition performance is often compromised by complex and variable road environments, dynamic lighting conditions, diverse designs of traffic signs, and real-time requirements. This paper is based on the YOLOv8 algorithm and utilizes the TT100K dataset for training and testing traffic sign recognition. The results demonstrate that the YOLOv8 algorithm exhibits excellent performance in traffic sign recognition. It achieves satisfactory recognition results for small traffic signs, traffic signs in complex backgrounds, and traffic signs under uneven illumination backgrounds. This lays a technical verification foundation for subsequent optimization of traffic sign recognition algorithms and research on multi-scenario applications.