基于改进YOLOv5的疲劳驾驶识别技术研究

Research on Sleep-deprived driving Recognition Technology Based on Improved YOLOv5

疲劳驾驶是导致交通事故产生的主要原因之一,也是目前各大汽车厂商在智能驾驶安全领域的研究重点。由于驾驶员面部疲劳特征凸显的时间特性离散且特征稀疏,因此研究提出基于YOLOv5算法提取主干网络浅层特征,并进行多特征融合,增设浅层特征预测头组成多预测头检测层,同时将轻量ECA注意力模块融入到Neck特征增强网络,共同组成改进的M-YOLOXs疲劳驾驶识别算法。通过实验分析统计表明改进算法相比较于原YOLOv5框架检测精度提高了2.68%,有效地提高了驾驶员疲劳状态的识别准确率。

Sleep-deprived driving is one of the main causes of traffic accidents,and it is also the research focus of major automobile manufacturers in the field of intelligent and safe driving.Due to the discrete and sparse temporal characteristics of the prominent facial fatigue features of drivers,a study proposes to extract shallow features from the backbone network using the YOLOv5 algorithm and perform multi feature fusion.A shallow feature prediction head is added to form a multi prediction head detection layer,and a lightweight ECA attention module is integrated into the Neck feature enhancement network to jointly form an improved M-YOLOXs Sleep-deprived driving detection algorithm,Through experimental analysis and statistical analysis,the improved algorithm has improved the detection accuracy by 2.68%compared to the original YOLOv5 framework,effectively improving the recognition accuracy of driver fatigue status.