Research On Technology Of Driving Fatigue Detection Based On Eye Features

The fatigue driving is one of the main reasons of traffic accidents, it becomes a research focus of transportation department all over the world, that how to detect drivers' fatigue state timely and effectively and reduce the traffic accidents due to fatigue driving. On-board, real-time, non-contact detection of driving fatigue in China and abroad remains one of the unsolved problems as yet, although there are some simple driving fatigue detections at present.This paper researches the principle of detection of driving fatigue, makes a comprehensive analysis and contrast to existing detection system, detailed analysis the key technologies and difficulties of several detect methods, and then puts forward driving fatigue detection based on eye features. Under the natural photoperiod, acquire video frequency under simulation driving, determine facial region utilizing face Gauss model, figure out eye region by color edge detection, and advance that recognizing Eyelid motion parameters by the ratio of eye's length to width and acquire Visual line parameters utilizes the relationship between gazing direction and iris position which confirm by Point Hough transform. At last, send the parameters into fuzzy logic to get the fatigue determinant result. For raising systematic efficiency, tracking eyes by combining Kalman filter with Mean Shift algorithm based on color histogram eigenvalue.We develop correlation algorithms of driver fatigue detection system by VC++6.0, and validate by simulation driving images, accuracy obtain 98% under general condition and drop to 89% when head rotating. The result shows that, this means could identify the fatigue statues of the tested person, and satisfies real-time control requirement, the detection results are more reliable and accurate relative to single eye feature.