Research Of Fatigue Detection Method For Drivers With Sunglasses

The increasing of vehicles not only brought convenience to people's life,also cause the threat to people's personal security.Fatigue driving is one of the important cause of malignant accidents,therefore,the study of detection method is of great significance to ensure the safety of people.However,the traditional detection method which is based on computer vision often cannot handle the drivers with sunglasses,which is lack of research.This method is based on the current situation and puts forward a kind of driver fatigue detection method which can deal with drivers with sunglasses.When drivers wear sunglasses,areas covered by sunglasses lose the color of skin so that the skin proportion of the eye area is reduced.Based on the basic principle,we can count the skin pixels distribution,form a skin color model and adjust it.It can determine whether a driver wears sunglasses.For the data which is not easy for color model method to distinguish,we can use method based on the deep learning to classify.Caused by a lack of the number of data,The method uses rotation,partial truncation,manual adjustment and elastic distortion to extend the data set to train the model for a better performance..The combination of the above two methods will make the error rate of sunglasses detection to a lower level.In the night or with sunglasses,we need to turn the infrared light on.And for sunglasses,we also need dodging processing.All of this is for the accuracy of the height of eyes and the judgment of the PERCLOS theory.Experimental results show that the given method can correctly classify whether drivers wear sunglasses and extract height of eye.The method has higher accuracy for driver fatigue detection and higher robustness to environment.