基于YOLOv5模型的驾驶疲劳研究

Research on Driving Fatigue Based on YOLOv5 Model

打哈欠是判定驾驶员疲劳状态的关键因素,考虑到驾驶疲劳检测易受驾驶员自身条件和外部环境的干扰,以及实时性差等情况研究了驾驶员的打哈欠问题,提出了一种利用YOLOv5网络模型进行疲劳检测的方法。首先通过LabelImg对处理过的YawDD开源数据集进行标注,再经过深度学习模型对样本进行多次迭代训练得到最优权重数据,最后将其用于测试集上进行测试。检测结果表明,样本平均识别准确率可达98%以上,所建模型具有高精度检测打哈欠行为的能力。

Yawning is a key factor in determining the state of driver fatigue,and considering that driving fatigue detection is susceptible to interference from the car driver's own conditions and external environment and poor real-time performance,this paper studies the driver's yawning and proposes a method for fatigue detection using the YOLOv5 network model.The processed YawDD open source dataset is firstly labeled by LabelImg,and then the samples are trained by deep learning model for several iterations to obtain the optimal weight data,and finally they are used on the test set for testing.The testing results show that the average recognition accuracy of the samples can reach more than 98%,so the used model has the ability to detect yawning behavior with high accuracy.