基于自适应更新背景的机车司机工作状态检测

Operation Mode Detection of Locomotive Driver Based on the Adaptive Updating Background

提出用追踪机车司机头部运动轨迹判别其工作状态。以改进的自适应混合高斯模型实现更新。一种光照下各个像素的模型均值构成一个均值矩阵。构建与每个均值矩阵相对应的固定背景宏块的高斯分布模型。对机车司机工作中的二值差分图像进行形态学处理,计算得到司机头部图像重心。在检测窗内跟踪机车司机头部图像重心的运动轨迹。用司机工作时的头部图像重心间的欧式距离辨识其工作状态。试验证明,此方法对机车司机工作中出现的疲劳发呆、反应迟钝或睡觉等非正常状态检测很有效。

This paper presents an approach to judge the operation mode of locomotive driver by using the movement locus of the driver＇s head. A modified mixture adaptive Gaussian model is used to update the background. A mean value matrix is composed of the mean value of every pixel of a frame background picture under the same illuminance. The Gaussian distributing model of fixed background micro block has been proposed corresponding to every mean value matrix. The driver＇s head barycenter is calculated by using morphology to process the binary difference driver＇s working picture. The locus of the driver＇s head barycenter is followed by detection window and the driver＇s working state is identified by the vary Euler distance of its head barycenter. Experiments on the railway show this approach is effective to detect the abnormality operation modes as being sleep or being in a daze.