摘要
针对现有疲劳驾驶检测报警不及时、检测精度不高以及需要人为监管的问题,提出一种改进YOLOv5s的疲劳驾驶目标检测算法。该算法使用轻量的EfficientNet骨干网络作为YOLOv5s的主干网络来进行特征提取,使模型参数大幅减少,降低模型的训练时间;同时选用SIoU作为模型的损失函数,优化模型损失计算方法,提升模型的检测精度。结果表明,优化后的YOLOv5s目标检测算法与原YOLOv5s相比,模型尺寸减少了2%,平均准确率提升了0.9%,能够有效提升矿用生产车疲劳驾驶目标的检测效果。
The alarms of the existing fatigue driving detections are not timely,the detection accuracy is not high,and the detections are highly dependent on artificial supervision,so a fatigue driving object detection algorithm based on improved YOLOv5s is proposed.In the algorithm,the lightweight EfficientNet backbone network is used as that of YOLOv5s for feature extraction,which reduces the model parameters and the model training time greatly.The SIoU is selected as the loss function of the model to optimize the model loss calculation method and improve the detection accuracy of the model.The results show that the model size of the object detection algorithm based on the optimized YOLOv5s is reduced by 2%and its accuracy rate is improved by 0.9%in comparison with the object detection algorithm based on the original YOLOv5s,so the proposed algorithm can effectively improve the effect of the fatigue driving object detection of mining production vehicles.
作者
杜威
宁武
孟丽囡
陈雨潼
DU Wei;NING Wu;MENG Linan;CHEN Yutong(School of Electronics and Information Engineering,Liaoning University of Technology,Jinzhou 121001,China)
出处
《现代电子技术》
北大核心
2025年第7期126-131,共6页
Modern Electronics Technique
基金
辽宁省教育厅基本科研项目(LJKFZ20220238)
