Research On Driver Fatigue Detection Method Based On Embedded

Fatigue driving has become one of the three major causes of serious traffic accidents.In serious traffic accidents,the proportion of accidents caused by fatigue driving is as high as 40%,so the harm of fatigue driving cannot be ignored.In this paper,the driver fatigue detection method based on vision is studied deeply,and a driver fatigue detection method based on embedded platform has been proposed.The main task of this paper is as follows:Ⅰ.Detection of driver’s face image.The Adaboost-based face detection algorithm has low accuracy under the condition of face occlusion and head posture change.In view of this deficiency,PCN algorithm is introduced into fatigue driving.PCN algorithm realizes accurate face detection by gradually correcting faces at different angles.The driver’s face image can still be accurately detected when the head posture deflection is more than 45 degrees.Tests on YawDD data set show that the accuracy rate of PCN algorithm for face detection is 98.91%,which is higher than Adaboost algorithm,and the tests provide data for later fatigue feature extraction.Ⅱ.Fatigue feature extraction.The feature extraction method of "three courts and five eyes" has large error under the condition of illumination,wearing ornaments and face occlusion.To solve this problem,SDM algorithm is used to locate facial feature points and extract fatigue features.Testing in a laboratory environment shows that SDM algorithm can locate facial feature points accurately.POSIT algorithm is used to estimate the head pose and obtain the head features of driver.The experimental results show that POSIT algorithm can accurately extract the driver’s head posture and ensure the accuracy of subsequent fatigue judgment.Ⅲ.Fatigue index test.In this paper,fuzzy reasoning is used to judge the fatigue degree.The triangular membership function and trapezoidal membership function are used to fuzzify the features.According to the physiological characteristics of fatigue,a fuzzy rule is designed that is more consistent with the actual fatigue performance.The fuzzy output function is calculated.Finally,the output function is deblurred to obtain fuzzy output.Fatigue judgment was manually marked on YawDD data set and 119 videos collected by the laboratory,and experiments were carried out.The experimental results show that the fatigue judgment accuracy rate reaches 93%,in this paper,the fatigue judgment method of fuzzy reasoning based on the three fatigue characteristics of eye opening degree,mouth opening degree and head posture can be used for driver fatigue driving detection.Ⅳ.Embedded transplantation.In this paper,iTop4412 development board is used,Linux system is installed,Qt is used to develop human-computer interaction interface,and Opencv library is used to the basic functions of image processing.The fatigue driving detection system designed in this paper is transplanted to the development board and tested in laboratory environment and vehicle environment.The experimental results show that the fatigue detection accuracy of the embedded fatigue driving detection system is 91.6%,and the algorithm takes 230 ms.The average time consumption of the algorithm is reduced to 170 ms when using multithreading program,which can be used for vehicle fatigue driving detection basically.