Research On Driver's Fatigue Feature Extraction Method And Design Of Detection System

In the function of economic and technology development,more and more office workers use cars as a means of transportation,which makes it easier to travel and save time.However,the accident on the road along with the widespread use of the car also appeared in the blowout growth.According to the investigation and scientific statistics,in the occurrence of traffic accidents for various reasons,fatigue driving as the main reason.Nevertheless,due to the immaturity and imperfect processing technology and software processing,products that detect the driver's fatigue can not be widely cited,thus failing to reduce or even eliminate the occurrence of similar traffic accidents,resulting in incalculable life loss of property.On the basis of analyzing the fatigue state of the driver,this paper extracts the fatigue facial information,combines the extracted multi-feature to determine the driver's fatigue state,and in the designed fatigue driving detection system verify the test results.Fatigue driving detection system includes three parts: face detection,fatigue feature extraction and fatigue state judgment.Among them,fatigue feature extraction includes eye features,head features and mouth feature extraction.In this paper,we first analyze the requirements and work flow of the system,and then discuss the algorithms used in each part.At the beginning of the test,this paper first used the Adaboost,which used after calculating the face Haar feature for face detection.In the use of the algorithm,I found that its speed and tilt the face of the detection have many deficiencies.Therefore,in the late,I introduced the KCF tracking algorithm,and combined them.In the application of contrast,this improvement makes the accuracy of face detection improved obviously,and shorten the detection time;then,the methods of eye feature extraction,head feature extraction and mouth feature extraction are discussed: In the extraction of the eye state,first the eye' sub-window is first determined on the priori knowledge of its' position,after the binarization process,the single eye window is extracted by using the gray integral projection,and the eye state is determined by calculating the external rectangle area;in the extraction of the mouth state,the nostrils and mouth regions are extracted according to the prior knowledge,and for the shortcomings of the traditional Canny edge detection,we used adaptive edge detection algorithm to obtain the contours of the mouth and nose,according to the distance of nostrils between the upper and lower lips to determine the state of the mouth;The determination of the head feature is based on changes in the coordinates of the rectangular box.After the fatigue feature extraction algorithms are discussed in this paper,introduced the method of fatigue state determination,for the lack of using single feature to exist fatigue detection,the method of combining the eye features,the mouth features and the head features is proposed,this method improves the accuracy of detection.In this paper,the fatigue driving detection system is designed in the laboratory through the VS development environment,and the fatigue detection method based on single feature fatigue detection and multi feature fusion is verified.The results show that the proposed method in this paper--the multi-feature fatigue detection method has a significant improvement in the accuracy.