Research On Driver Fatigue Detection Technology Based On Video Sequence

With the rapid development of transportation in China,there is no doubt that the emergence of cars provides a lot of convenience for people’s travel,but a series of safety problems follow.Every year,there are hundreds of thousands of traffic accidents around people.Behind many traffic accidents,drivers often hide the fact that they are tired driving.The seriousness of fatigue driving is self-evident,how to solve this problem has gradually become the focus of social attention.In this thesis,the driver fatigue detection technology based on video sequence,do some pretreatment before fatigue test,in the process of fatigue testing,according to the positioning of the order by the driver’s face the eyes,to get the eye area and identify the states of eyes,through calculating the related parameters of fatigue of eyes,to determine the condition of the driver.The main research contents of this thesis are as follows:(1)In order to solve the problem of more noise and uneven illumination caused by the acquisition equipment itself and illumination Angle in the process of image acquisition,adaptive median filtering is adopted to denoise and illumination compensation is carried out through multi-scale self-quotient graph to ensure the smooth follow-up driver face detection.A face detection method is proposed,which uses Cb and Cr clustering features to detect skin color,mark the image,and remove the neck interference based on pixel gray center line to complete face detection of the driver.(2)The eye detection methods based on integral projection are studied and contrasted,including integral projection method,gradient integral projection method and improved gradient integral projection method.Aiming at the disadvantages of the traditional integral projection method,which is easy to be affected by light and has low localization accuracy,this thesis analyzes the characteristics of the gray distribution of eyes in the horizontal direction,and uses two line gradient operators to improve the integral projection method.The human eye positioning process is divided into two steps,the left eye and the right eye are segmented.(3)In order to improve the speed of human eye state recognition,the aspect ratio of black pixels in human eye region is adopted.By calculating the number of horizontal black pixels and vertical black pixels in the human eye area,the ratio of the two is converted to the information of eye opening and closing,and the driver fatigue is determined by calculating PERCLOS value and blink frequency.(4)A system is designed on PC to simulate fatigue detection,and the whole detection process is displayed by graphical user interface.The final experimental results show that the system can effectively detect the driver fatigue state,which has a certain practical significance.