Dou Huili, Chen Sirui, Xu Fangyuan, Liu Yuanyuan, Zhao Hongyang
Zhejiang Institute of Communications, Hangzhou, Zhejiang, China.
PLoS One. 2025 Mar 18;20(3):e0314817. doi: 10.1371/journal.pone.0314817. eCollection 2025.
This study proposes an improved YOLOv8 model for vehicle and pedestrian detection in urban traffic monitoring systems. In order to improve the detection performance of the model, we introduced a multi-scale feature fusion module and an improved non-maximum suppression (NMS) algorithm based on the YOLOv8 model. The multi-scale feature fusion module enhances the model's detection ability for targets of different sizes by combining feature maps of different scales; the improved non-maximum suppression algorithm effectively reduces repeated detection and missed detection by optimizing the screening process of candidate boxes. Experimental results show that the improved YOLOv8 model exhibits excellent detection performance on the VisDrone2019 dataset, and outperforms other classic target detection models and the baseline YOLOv8 model in key indicators such as precision, recall, F1 score, and mean average precision (mAP). In addition, through visual analysis, our method demonstrates strong target detection capabilities in complex urban traffic environments, and can accurately identify and label targets of multiple categories. Finally, these results prove the effectiveness and superiority of the improved YOLOv8 model, providing reliable technical support for urban traffic monitoring systems.
本研究提出了一种改进的YOLOv8模型，用于城市交通监测系统中的车辆和行人检测。为了提高模型的检测性能，我们在YOLOv8模型的基础上引入了多尺度特征融合模块和改进的非极大值抑制（NMS）算法。多尺度特征融合模块通过组合不同尺度的特征图，增强了模型对不同大小目标的检测能力；改进的非极大值抑制算法通过优化候选框的筛选过程，有效减少了重复检测和漏检。实验结果表明，改进后的YOLOv8模型在VisDrone2019数据集上表现出优异的检测性能，在精度、召回率、F1分数和平均精度均值（mAP）等关键指标上优于其他经典目标检测模型和基线YOLOv8模型。此外，通过视觉分析，我们的方法在复杂的城市交通环境中展示了强大的目标检测能力，能够准确识别和标记多类目标。最后，这些结果证明了改进后的YOLOv8模型的有效性和优越性，为城市交通监测系统提供了可靠的技术支持。
