Research Of Driver's Fatigue State Detection Method Based On Video

Drivering fatigue is a major factor for traffic accidents.So,how to effectivly monitor and prevent of drivering fatigue is significant to reduce traffic accident.In the course of driving,whether fatigue or not can be reflected in driver eyes open degree,the use of their eyes open degree to determine status of driver fatigue is an efficent solution. How to detect the eyes' open degree fast,real-time,accurately is the main research objective in this paper.The paper includes the following three aspects:Firstly,under the special circumstance of the operator cabin,the motion object including the driver's face is detected by the method of background subtraction.The illuminance variation which affects the detection of the motion target is mainly considered in this paper.A new background updating method is proposed to solve the problem of the illuminance changing which takes place slowly or quickly.The face is located by ellipse fitting,and then the result is checked by the face's intensity character. The method described above can locate the face concisely,quickly and accurately.Secondly,According to the result of face location,symmetry transform is used to search the glabellum and locate the eyes roughly.Then the method of integral projection is used to accurately locate the driver's eyes.In the case of the driver wearing glasses, the eyes semi-open,complex background existing and the head rotating,the method in this paper can locate the eyes quickly and accurately and the detection accuracy is high.Thirdly,Hough transform and Daugman operator are used to extract the outline of the eyeball,then,the ratio between the area of the eyeball and the area in the outline is computed,the ratio can show the open degree of eyes.Finally,the proposed algorithm is combined with the POFECT method to decide whether the driver is fatigued.The algorithm's simulation and relevant experiments are processed in the MATLAB environment.Experimental results show that the predetermined target is achieved and the algorithm presented in this paper is effective.