| With the acceleration of the pace of social development,people's quality of life has greatly improved,and residents' car ownership has increased year by year.In automobile driving,besides drunk driving,fatigue driving gradually becomes another important factor causing traffic accidents.Therefore,real-time fatigue detection for drivers has far-reaching significance for the prevention of major traffic accidents.This article through reading a large number of literature and information,understand the current development status of fatigue detection methods at home and abroad,and based on this study in-depth study of multi-feature fuse driver fatigue detection.The key technologies in fatigue detection mainly include: driver's face detection and tracking,the location of fatigue area in facial subspace,and the extraction and fusion of fatigue features.Firstly,in order to reduce the influence of the outside world's complex environment on the driver as much as possible,this paper converts the captured video into frames,and then preprocesses the images,which mainly include the graying,equalization and adaptive illumination compensation of the image;By comparing color in different color space clustering features,YCgCr skin color threshold segmentation is selected for face detection,and the detected face image is input into the improved Viola-Jones framework,combining two kinds of face detection algorithm to realize the face region.The single-frame image does not reflect whether the driver is in a state of fatigue.Therefore,the Camshift algorithm is used to track the face video sequence to meet the real-time requirements of the system.Secondly,based on the detection of the human face,the positioning of the facial fatigue deformation region is achieved.The fatigue region mainly includes the eyes and the mouth.In the positioning of the eyes,the method of combining the priori knowledge of the geometrical distribution of the face and the integrated gray-scale projection of the driver's image and the detected face region respectively separates the eyebrows and achieves coarse positioning of the eye;By comparing the detection effects of several edge operators,Sobel edge operator is finally selected to achieve the extraction of eye contours.The distribution of the human face conforms to the geometrical laws of "three-room and five-eyes",and the mouth area is located according to the relative positions of the human eye and the mouth.Finally,the fatigue state detection of multi-feature fusion is mainly achieved.Extracting PERCLOS parameters from the already located eye area and extracting the YawnFreq parameter from the mouth area;According to the threshold of two parameters of human fatigue,its fuzzy set and universe are determined,and the expression of membership function is obtained through calculation and derivation;building a fuzzy logic inference system and two fatigue operators The input is integrated into the fuzzy system,and the driver's fatigue state is output.The fatigue degree FatigueLevel is obtained by anti-fuzzification to realize the driver's fatigue state recognition. |
PDF Full Text Request