Estimated Driver Fatigue Detection, Eye Detection And Line Of Sight

The accidents caused by drivers' fatigue take a great part of all the motor vehicle accidents. When the driver is in fatigue, we can give him an alarm to prevent him from fatigue. It is very useful to reduce the accidents. And the states of the drivers' eyes could reflect whether the drivers are in fatigue. It is a feasible way to judge whether the driver is in fatigue by the states of his eyes. In this dissertation, we give a non-contact algorithm based on computer vision to detect the drivers' fatigue. We get the images of the driver by a normal camera, and then we use the algorithm in this dissertation to detect the driver's fatigue by analyzing the state of eyes in the images.The driver fatigue can be divided into two types. One is the micro-sleep and another is the scatterbrained. First, we locate the driver's eyes by the gray feature of the eyes. In the algorithm of the eyes detection, we use the edge detection algorithm, the gray projection algorithm and the template matching algorithm. And based on locating the iris, we detect the state of micro-sleep. Secondly, the SUSAN corner detection is used to detect the eyes' corners. To improve the speed and the precision of the eyes' corners detection, we restrict the searching area by the location of iris. By the iris' center and the eyes' corners, we can estimate the direction of the line of sight. In the end, we use dynamic Bayesian network to fuse the state of micro-sleep and the state of scatterbrained. Considering that the fatigue detection is related with the prior knowledge and probabilistic statistics, the dynamic Bayesian network is used as the analysis tool to perform the reasoning of fatigue. By the studying of the fatigue mechanism, we get the network configuration. And by counting a lot of samples, we can get the parameters of the dynamic Bayesian network. Experiments show that in the most of situations the proposed system works and the corresponding performance is satisfying.