Research And Implementation Of Eye Detection And Tracking Algorithm To Fatigue Driving Detection

As the swift development of the economy and the rapid growth of vehicle population in the world, the traffic safety problem has become the global hot issues. According to a survey, the fatigue driving is one of the major reasons why traffic accidents happen. Therefore, a fast, accurate and real-time fatigue driving detection method has become an important means to reduce these traffic accidents. At present, most detection methods are to estimate fatigue level by eyes’status, however, the precondition of judging the eye status is to detect the position of the eyes. Therefore, eye detection and tracking algorithm is a crucial step in the fatigue driving detection method.After analyzing the traditional algorithms, this thesis researches and implements eye detection and tracking methods to fatigue driving detection. First of all, median filter and non-linear transformation are adopted as image preprocessing in order to eliminate the influent of noise and uneven illumination for the subsequent operations. Secondly, in order to speed the eye detection, face location method based on skin-color model is used to narrow the scope of the eye to the face region. Eye detection method based on isophotes is adopted after positioning the face. This method applies the object detection based on isophotes to eye location, is independent on the shape and size of eyes, and is steady for image-whirling in any direction. Finally, eye tracking is taken for increasing the speed of eye detection. A novel eye tracking method based on area-block pairing and template matching is presented. The whole eye tracking method divided into three stages:estimate eye search scope through analyzing movement law of the image sequence; obtain eye candidate regions according to the position of the eyes in face region, eye features and the geometric relations between the eyes; locate the position of the eyes by template matching to achieve eye tracking. This eye tracking method needs less calculating amounts, costs less time, has excellent robustness to the driver’s head rolls and is effective for normal lighting condition and eye closure.Firstly, this thesis introduces the research current situation of eye detection and tracking to fatigue driving detection. Secondly, the processes of the face location based on skin-color model and the eye detection based on isophotes are elaborated. Then, this thesis presents a novel eye tracking method based on area-block pairing and template matching. At last, experiments are carried out to evaluate the algorithm performance under the platform of Matlab R2008a. The experimental results show that the eye detection and tracking method presented in this thesis has higher correct detection rates, lower false detection rates and lost rates, can meet the real-time demands of the fatigue driving detection.