Study On The Detection Of Fatigue Driving Based On Characteristics Of Eye Movement

With the rapid development of the social economy, people's consumption capacity increased year by year, a sharp increase in the number of cars has brought more security risks to the entire transportation system, Fatigue driving has become one of the main factors which cause traffic major accidents. It is effective to avoid the occurrence of traffic accidents if the drivers can be in time to avoid the traffic accidents, it has a great significance for traffic safety.This paper is based on a series of image processing methods and algorithms based on the characteristics of the eye state of the driver, mainly include face detection, eye location, eye state identification, fatigue state detection of the four major processes, by identifying the eye state and using the P80 standard of PERCLOS to detect the fatigue state. The specific work is as follows:(1) This paper elaborates the driver face detection method based on Adaboost algorithm, including Haar feature selection and extraction, the calculation of integral figure, the training of the classifier. The classifier is trained by extracting the Haar_like features of the samples, all the weak classifiers trained are used to form a strong classifier, then use the strong classifier for face detection and positioning.Due to the large number of samples, in order to improve the speed of detection and reduce the search range of the strong classifier, to determine the true location of every weak classifier. and then to detect the face region.(2) This paper studies the method of human eye localization based on geometric features. According to the geometric distribution of the eyes in the face,and the distribution of connected domain after binarization to calculate the centroid of each connected domain, and to determine the coordinates of the centroid of the human eye, and according to the centroid's coordinates of the human eye by bilinear interpolation method of inclination larger face image rectify and to determine again the centroid's coordinates of the human eye, about the interception of any one area of the eye diagram.(3) In this paper, the human eye state recognition method based on PCA principal component analysis and gray level projection method is presented in this paper. PCA algorithm is adopted to find the most representative feature space of human, contrast sample images and test distance vector of images, use vector difference as recognition criteria; gray projection method is to determine the height of the pupil by the human eyes' grayscale distribution, namely, to determine the stateof the eye by calculating the pupil's height of the human eye of the single frame image. PCA is to extract maximum grey value and minimum gray value while human eyes' state is open or close, and then use gray projection method to calculate the human eye opening, Finally, to identify the state is by the fatigue criterion.(4) In driver fatigue detection process, using PERCLOS standard to count the percentage of eye continued closure time of a specific period of time, and P80 is selected as the evaluation index. The experimental results show that the fatigue detection method used in this paper has good reliability and real-time.