The Research Of Fatigue Driving Detection Based On Image Processing

As the traffic accident caused by fatigue driving has increased fast,The detection of thedriver fatigue has been more important.At present,Domestic and foreign experts havedeveloped a variety of fatigue driving detection method.Such as heart rate tester,lanedeparture warning system,head position detection system,steering wheel detection system andso on.Because of the limit of the driving environment,some detection systems which weredirect contact with the driver have an impact on the driver,Combined with their advantagesand disadvantages,eye could reflect the fatigue state of the driver directly,so in this paper,theresearch of the fatigue driving detection is based on the combination of PERCLOS criterionand blink frequency in the image processing,this method is one of the hot research in the fieldof fatigue driving detection.The detection of the fatigue driving was based on the knowledge of image processing,itdetected the fatigue driving mainly based on the fatigue characteristics of the driver,Thewhole system determined the driving state by driving PERCLOS value and the frequency ofblinking of the eye region.In the detection, it included pretreatment of the picture, facepositioning, human eye positioning,detecting of the state of the human eye, PERCLOS valueand blink detection frequency.The face location was detected by combining the YCbCr colorfeatures and a small area removed, holes filled and jaw tracking algorithm.Improvedpositioning algorithm exclude the false detection of skin area of the neck,it detected the faceregion more accurate;Eye location was detected by combining the gray integral projection andimproved template matching method,it not only made the positioning more accurate,but alsoreduced the complexity of the traditional template matching detection algorithm,Then thedetection efficiency was improved;In the eye state extraction, the paper was chosen the grayarea of comparative law, the algorithm was simple and robust better.Finally,in the fatiguedriving determination,it was detected by the method of PERCLOS criteria and blinkfrequency.The driver fatigue driving simulation was running in VC++6.0environment, theaccuracy rate of the driver fatigue detection system was about90%,the detection efficiencywas improved.Experimental results show that: The computational complexity is reduced greatly by theimproved face location, eye test, eye state extraction algorithm,and it has a higher accuracy.Itcontents real-time and accuracy in the fatigue driving determination. So the proposed driver fatigue detection system has high practical value.