Driver's eye state detecting method design based on eye geometry feature

Abstract

When we use the computer vision to inspect the driver's driving behavior, the identifying of eye states is one of the key technologies. In fact, when a driver drives in a normal, doze or sleeping state, his/her eye opening degree will quite different. According to this phenomenon, this paper uses the eye regional geometry characters as the feature values, and they are formed into an eigenvector as the input to a three-level BP net. Then we can get the output of three different spirit states through the neural net. The experiment results show that this new method can inspect the driver's eye states rapidly and effectively.