Study On The Detection Of Fatigue Driving Based On Image Processing

Fatigue driving is one of the causes of traffic accidents. The driver fatigue state detection and alarm can effectively reduce the possibility of traffic accident. Many of the current researches determine the status of driver fatigue through the physiological characteristics, detection of the driver’s head displacement, body movement characteristics. The fatigue driving early warning system based on machine vision that use PERCLOS algorithm on driver’s fatigue state detection can effectively detect driver fatigue and give early warning information.This thesis’core is the image processing algorithm, including a series of steps of image processing, and the time complexity of the algorithm should be considered to meet the requirement of real-time processing while comparing and selecting each step. Firstly, according to the characteristic that the car image is vulnerable to noise interference, the image is median filtered, so as to eliminate the influence of noise. Then, according to the characteristic that the car image illumination condition is unstable, equalize the image histogram to avoid the image too bright or too dark.First locate face to narrow the search scope of eye location and improve the detection efficiency. After locating the face through the skin color segmentation, eliminate non-face region and get the exact position and size of face. In the premise of both real time and accuracy, use the Adaboost algorithm to locate eyes, use the gray projection algorithm to detect eyes status, and the results will be used as parameters during PERCLOS computing. Then, calculate the PERCLOS value during a certain period of time (10s in this paper), and compare it with the standard threshold, to judge whether the driver is in a state of fatigue.This paper designs the corresponding hardware system, captures the image information of driver by the CMOS camera and transmits it to the car PC through AVR controller. Through the tests for different driving video, the system is able to locate the driver’s eyes and analyse the eye states. The experimental results show that the PERCLOS value exceeds the threshold while the driver is fatigued, and lower than the threshold while the driver is fatigued. Thus, the system is able to judge that whether the driver is fatigued.