A driving assistant safety method based on human eye fatigue detection

Driver's Fatigue is one important factor leading to traffic accident. So, how to detect driver's fatigue status accurately is critical to avoid traffic injuries and deaths. In the driving environment of vehicles, there are higher requirements for the real-time performance and reliability. The traditional fatigue judging methods are sensitive to the light fluctuation and have lower real-time performance. This paper improves eye location method and presents a fatigue judgment rules to meet the demanding of the automobile driving environment. In the whole process, we take inter-frame information to define eye region to improve the eyes location method based on AdaBoost. Then, we combine skin color segmentation and eyelids ellipse fitting method to judge the eye state. In the last, we design a fatigue recognition rules based on rough set which combines closing eyes rate, average eyes blinking time and consecutive eye closure frames together. These methods can achieve better real-time performance. The experimental results show this method can meet with the requirements of driving environment and improves fatigue judgment accuracy rate.