Research On Fatigue Detection Method For Train Drivers Based On Facial Multiple Feature Fusion

With the improvement of railway technology of China,the operation method of train drivers which continuously monitors train operation information in real time has replaced the original train control method.In addition,combined with the characteristics of prolonged running time and complex road conditions,train drivers burden more and more workload so as to increase the possibility of fatigue driving.Therefore,it is significant to detect train drivers’ fatigue in real-time and give early warning to ensure the safe operation of the train.In order to improve the dependability and correctness of the fatigue detection system for train drivers,the image brightness adjustment method of block processing is adopted to reduce the influence of illumination changes on the fatigue detection system.Aiming at the problem that the fixed threshold cannot be applied to each driver due to the difference of different drivers’ facial features,the adaptive threshold is formulated according to the facial features of each driver,which integrates various facial features through fuzzy reasoning to realize the online detection of train drivers’ fatigue driving.The main research contents include:(1)An image brightness adjustment method based on block-processing is adopted,which uses different image processing methods in accordance with different image brightness conditions to ameliorate the influence of uneven illumination or too dark and too bright on the fatigue detection system.(2)Comprehensively considering the correctness and real-time requirements of fatigue detection,the face is detected by the face detection method based on HOG features,and the opening and closing degree of the eyes and mouth are calculated by the feature point location method based on ERT.The pupil location is realized,and the eye movement rate is calculated by combining the obtained pixels with the face feature points.(3)In order to make the threshold of fatigue characteristic parameters suitable for every driver,an adaptive threshold algorithm based on k-means++ is used.The eye opening and closing degrees of every driver in the fatigue cycle are clustered into two categories by k-means++ method to judge the eye state and make the selection of threshold adaptive.So as to further distinguish the yawning and other mouth opening behaviors of train drivers,the public data set is used to detect the mouth in the natural state of the mouth,speaking and yawning.It is divided into 3 categories by k-means clustering,and thus the threshold value in the case of yawning is worked out.(4)The accuracy of fatigue detection which is up to 95% is improved by fusing the three indicators of eye opening and closing degree,mouth opening and closing degree and eye movement rate through fuzzy reasoning system,and the fatigue classification which meets real-time requirements is realized.