| With the development of road traffic,the issue of road traffic safety has been attracted more and more attention,and the phenomenon of driver fatigue detection is becoming increasingly serious,which is one of the major causes of traffic accidents.Researching the real-time driver fatigue detection system to make reminders and early warnings when the driver is fatigued is of great significance for reducing traffic accidents caused by driver fatigue detection.This paper studies the driver fatigue detection technology based on machine vision,and divides the process of driver fatigue detection into three steps: face detection,face shape regression,fatigue feature extraction and recognition.The algorithm used in each part is studied in depth.At the same time,a fatigue detection program was developed on the ARM development board based on the theoretical research.Through the processing of the images from the webcam,the driver’s eyes were analyzed to determine whether the driver is currently fatigued.The main content of this paper is as follows:(1)Face detection.This paper studies the face detection algorithm based on Adaboost,and constructs a face detector based on OpenCV and open-source face database,and tests the trained face detector.(2)Face shape regression and eye positioning.This paper studies the random forest regression model.Then,based on the face detection,the feature point locator was trained and tested on the Hellen face database using the random forest regression model.(3)Fatigue feature extraction and fatigue state discrimination.Through the feature points around the human eyes,the fatigue characteristics are extracted by human eye opening and closing degree.In consideration of the shortage of the traditional methods of using PERCLOS to judge the fatigue state,a method based on CART decision tree to judge the fatigue state of drivers is proposed.Finally,the effect of these two methods was tested and compared.(4)Realization of driver fatigue detection system based on ARM development board.On an ARM development board equipped with an Android system,an image of a network camera is received via a LAN to implement a driver image acquisition system and a fatigue detection system.Finally,the tests on the fatigue detection system show that the system has high detection accuracy for fatigue driving and can give early warning,which meets the requirements of real-time fatigue detection. |
