Research On Fatigue Driving Detection Technology Based On Eye Information And Head Posture

In recent years,frequent traffic accidents caused by fatigue driving have caused extremely serious consequences for social property and life safety.Therefore,it is of great practical significance to research accurate and efficient fatigue driving detection technology to determine the driver’s fatigue state in advance and give early warning.The method of classifying and identifying the driver’s visual features by using computer vision technology has become an important method in the field of fatigue driving detection with the advantages of good real-time performance and non-contact.Based on the comprehensive consideration of the real-time performance,accuracy and environmental adaptability of the algorithm,this paper focuses on the face detection and face feature location,which restrict the accuracy and timeliness of the algorithm.Aiming at the problem that the facial features of the current fatigue detection method are easily affected by the face attitude Angle,light changes,individual differences and other factors in the practical application.Combined with the characteristics of fatigue driving test task to improve it.This paper presents a fatigue driving detection method based on multi-features of eyes and head.The main research contents are as follows:(1)Aiming at the problem that the traditional face detection algorithm based on the Adaboost classifier is not high in real-time,this paper introduces difference hash fingerprint information to improve it.This paper adds a cache mechanism to store hash fingerprints and face positioning boxes of similar pictures.If the two frames of pictures are very similar,face detection is not required,but the face positioning box is directly called from the cache module.In this way,the recognition time of face detection is greatly reduced.(2)Aiming at the problem that the driver’s posture angle affects the positioning accuracy of facial feature points during driving,a facial feature point localization algorithm based on local binary features(LBF)is selected.And the initialization strategy of the algorithm is optimized.Instead of simply using the average face model for initialization,this paper uses the histogram of oriented gridients(HOG)combined with the support vector machine(SVM)algorithm and the convolutional neural network algorithm to determine the face orientation.The two methods are analyzed and compared,and different facial feature points are used to initialize the model according to the different results of the face orientation.Aiming at the problem that the algorithm is easily affected by factors such as illumination variation,a normalized feature is proposed to replace the original feature to establish LBF random forest.Thus,more feature information is retained,the classification ability is stronger and the computation amount is not increased significantly.(3)Aiming at the problem that the commonly used fatigue detection method based on visual features is single and has poor robustness,a fatigue detection method based on multiple features of eyes and head is proposed.Multi-characteristics such as the degree of eye closure,blink frequency and head posture angle were used to comprehensively determine the degree of fatigue.Aiming at the problem that the above-mentioned fatigue characteristics are not obvious when the driver is slightly fatigued,it is proposed to distract the driver’s attention as a kind of light fatigue index.Optimized pupillary center localization algorithm.Establish a line of sight estimation model.Determine the location of the fixation area to determine whether or not attention is being diverted.This allows drivers to be warned before they fall into severe fatigue.Finally,this paper integrates various modules through experiments and comprehensively evaluates system performance.The results show that this fatigue driving detection algorithm not only has high accuracy,but also has good real-time performance,and can effectively detect the fatigue state.