Research Of Fatigue Detection System Based On Human Eyes Detection

A large number of accidents occur every year, tens of thousands of accidents caused by driver fatigue, so how to reduce accidents caused by driver fatigue is extremely important.Driver fatigue detection technology which is based on visual characteristics has become a hot research because of its non-contact, non-interference and practicality.Fatigue detection is a method based on human visual facial features and head posture analysis. As the unique physical structure and perfect symmetry of human eyes,the eye detection and analysis of the state human eyes have been the most effective method for fatigue detection.We extract the face area,then take advantage of human eyes center location based on Dynamic Morphological Quotient model,combined with an unsupervised feature selection based on local reconstruction to get the precise position of the human eyes. And then analyze the state of human eyes by means of PATECP algorithm, then get the final determination of fatigue according to PERCLOS.Because of the main problem to be solved for the research status of the human eye detection, such as complexity of illumination, changes of head pose,the block of glasses, the algorithm accuracy and speed need to be improved.The main work is as follows:Firstly,morphology quotient image can make characteristics of person eyes to be outstanding,We utilize geometric relationship between eyes and lip proposing a geometric eye detection method based on DMQISecondly,due to the geometric eye detection method based on DMQI mainly using of geometric relationship of the eye and the mouth, robustness is not strong.In this paper,unsupervised feature selection based on local reconstruction can process large amounts of data,meanwhile human eyes and mouth can be outstanding by DMQI,combined with the Bayesian posterior probability,we propose a new human eyes center location based on DMQI model with a higher position accuracy, more adapted to the face posture,a variety of light conditions can be overcomed,at the same time,it has strong robustness.Finally,experiment results show that the human eye position is located accurately in the experimental conditions,the accuracy and real-time of the fatigue detection system has improved greatly.