一种基于列车司机在途疲劳状态检测及预警的方法
2021-05-09江跃龙



摘 要:针对列车司机肤色与背景图的聚类分布特性较好分离,而且具备高斯两维独立分布,可将列车司机图像分布区域与列车驾驶室内部背景图很好地分开,将列车司机的头部信息,作为人脸检测定位算法的原始训练集。反复修改训练列车司机图像数据的权值分布,最终将背景图像剔除掉,得到列车司机的人脸图像。确认人脸区域后进一步确认眼睛位置,实现级联结构增强的定位识别,在利用眼睛状态判定时,则用眼部和嘴部的融合状态判断驾驶员是否疲劳,相对提高了疲劳判断的全面性和准确性。
关键词:疲劳驾驶;智能联动检测;眼睛状态判断;人脸检测率
中图分类号:TP391.4 文献标识码:A文章编号:2096-4706(2021)20-0098-05
A Method of Detection and Early Warning Based on the Transit Fatigue
State of Train Driver
JIANG Yuelong
(Guangzhou Railway Polytechnic, Guangzhou 510430, China)
Abstract: For the clustering distribution characteristics of the train drivers skin color and the background image is easy to separate, and has Gaussian two-dimensional independent distribution. The distribution area of the train drivers image can be well separated from the background image inside the train cab. The head information of the train driver is used as the original training set of the face detection and location algorithm. Repeatedly modify and train the weight distribution of the data of the train driver image, and the background image is removed finally to get the face image of the train driver. After confirming the face area, further confirm the eye position to realize the location recognition enhanced by the cascade structure. When determining through the eye state, the fusion state of eye and mouth is used to judge whether the driver is tired, it relatively improves the comprehensiveness and accuracy of fatigue judgment.
Keywords: fatigue driving; intelligent linkage detection; judgment of eye state; face detection rate
0 引 言
在軌道交通运营过程中,由于整个城市轨道交通系统是错综复杂的,在列车行驶过程中如出现任何故障都会造成一定损失,为了有效地保障轨道交通运营的稳定性与可持续性,需要开展地铁列车司机的驾驶风险监督防护措施和维保工作,以提高列车行驶安全。虽然采用人工智能技术与计算机系统作为辅助,但作为高速轨道列车运行的直接指挥和操控岗位行车人员在工作过程中需要精力高度集中,这也将导致列车司机在驾驶列车过程中发生司机注意力不集中、打瞌睡等状况时,极易产生工作疲劳,容易造成安全事故[1]。操控岗位行车人员的工作行为、精神状态需要对列车司机或者行车人员的工作状态,如工作是否疲劳、有无错误操作。……
