Fatigue Driving Detection System Research Based On Multi-feature Information

As we know,fatigue driving has become "the number one killer" of traffic safety,and with the popularization of driver fatigue warning system, the continuous development and improvement of fatigue driving detection system technology is imminent. To overcome the limitations of a single sensor in fatigue detection, we improve the fatigue driving detection rate through comprehensive analysis these three features of the driver’s eyes and mouth status,and the vehicle traveling trajectories.First of all, this paper obtain the driver’s eyes, mouth and the road feature data from the driver’s facial image and road image through installing cameras in the car and the front of the car,using image acquisition program of the infrared light auxiliary sources to ensure the all-time availability of the system and filter out the effect of natural light;extracting the eye fatigue characteristic by texture analysis infrared detection method,and then detect and locate the eyes and mouth area by the intelligent ADABOOST algorithm; we first locate the nose,and then enhance the mouth detection rate through the geometric position relationship of nose and mouse,then calculate the opening degree of the mouth according to the area the size;calculating lane of road image using a simple linear lane model, and obtain the parametric of the linear lane model using the better robustness probability Hough transform, so calculate the yaw rate.This paper propose the eyes closed time rate to analyze eye feature, on the basis of PERCLOS algorithm, this method let the ratio of the detected eyes closed time and normal eyes closed time as the standard of fatigue analysis, for the mouth fatigue Characteristic Analysis, we use the dual-threshold yawn analysis methods to analyze fatigue and distinguish non-yawn; this paper detects lane deviation using the yaw rate, which is defined as the ratio of the left and right lane lines and the image horizontal axis angle. Finally, do fusion analysis using the weighted average, with select the PRECLOS, the eyes closed time, yawn time and the lane yaw rate as the features value, experiment show that the accuracy rate has greatly improved.