Research Of Fatigue Driving Monitoring Method Based On Face Recognition

The rapid development of the transportation industry to make car ownership and the number of accidents increased year by year. The fatigue driving has become one of the main causes of traffic accidents, so the research of fatigue driving monitoring methods has become an important topic. Fatigue driving monitoring based on face recognition technology has a good real-time, high accuracy, non-contact and strong robustness characteristics, to reduce the traffic accident caused by fatigue driving has the profound significance.The driver’s facial images as research object in this paper. Locate and extract the eyes and mouth images based on face recognition. Then identify the state of the eyes and mouth. Consider the variety of facial fatigue information to determine the extent of fatigue driving. The main research contents are:Firstly, study of face recognition. By summarizing the common face recognition algorithms and combining with the actual environment of the cab, we use Adaboost algorithm for face recognition. Face detection experiments were performed on images with facial rotation, complex background and actual collection.Secondly, research on the eyes and mouth location method. State of the eyes and mouth can be response to the degree of fatigue driving, so this paper uses the method of integral projection curve to locate and extract the eyes and mouth images. A picture of ORL facial image database as an example explained how to locate and extract the eyes and mouth images use integral projection method. Actual collection images are carried on the eyes and mouth location experiment.Then implement the eyes tracking. Eyes tracking based on eyes location in order to improve the quickness and real-time monitoring. Traditional particle filter algorithm due to fixed number of particles leaving the high time complexity. Using adaptive particle filter algorithm which can be dynamically adjusted the number of particle to solve this problem. And validate the algorithm in actual collection video images.Finally, complete the information extraction and discrimination of fatigue. Adopt the improved horizontal projection method to identify the state of the eyes and mouth. That is the eyes and mouth were opened or closed. Extract four kinds of the eyes and mouth fatigue information, respectively is the value of PERCLOS, BlinkFreq, ECT and the number of yawning. By considering four kinds of fatigue information to determine whether the driver is fatigue. Fatigue discrimination experiments were performed on video images with actual collection.In this paper, face recognition, the eyes and mouth location, the eyes tracking, fatigue information extraction and fatigue discriminate experiments on the actual collection images. Experimental results shown that fatigue driving monitoring based on face recognition can meet the requirements of quickness, accuracy and real-time. It can accurately monitor the state of fatigue and has high practicability.