Research On Driver Fatigue Detection Based On Eye Detection

Each year, fatigue driving leads to tens of thousands of traffic accidents and a large number of casualties around the world. In China, fatigue driving has even been listed as the one of the three major causes of road traffic accidents. How to timely and effectively detect the degree of a driver's fatigue and reduce the number of driving accidents caused by fatigue has become a research hotspot in current Intelligent Transportation System.Through the comparison of principles and methods of existing fatigue detection technologies at home and abroad, as well as analysis of advantages and disadvantages of different detection methods, taking into account the car, real-time, non-contact requirements, this paper chose to use PERCLOS method to build fatigue detection system. During the process in detection of the human eye, we introduce Haar-like feature to detect changes in gray degree in local characteristics of the human eye, through the integration plan to calculate characteristic values, improving the detection speed, and finally make the weak classifiers to form a strong classifier trough cascade to achieve accurate detection of the human eye. During the training process, the degradation, which is caused by too much emphasis on the difficulty samples, is resolved by tagging a identification in specific samples, and timely release such a sample weight and be normalized to alleviate the degradation.For the human eye state identification, based on the detailed analysis of the two basic algorithms, we propose the use of two sets of templates, open eyes and closed eyes, by standardization of size and gray distribution, we use template matching to determine the status of the human eye, while according to results of discriminant to calculate PERCLOS value, if the value exceeds the threshold driver is determined fatigue. In order to simplify the calculation, the ratio of the time in this article is converted into the ratio of successive frames.At last, this paper built the driver fatigue detection system, and the algorithm is verified. Experiments show that under laboratory conditions, the above method can accurately detect the eye region, and it can issue a warning according to real-time state of eyes'open and closed, so that it meets the needs of the driver fatigue detection.