A Study Of Driver Fatigue Detection Based On Iris Detection

Abstract:With the development of society and the improvement of people’s living standard, the car ownership has also increased, the traffic accidents has increased dramatically. While the proportion of fatigue causes is very high, so it is very important to prevent and supervise the driver fatigue driving, and that measures can reduce the number of traffic accidents. There has been a lot of researches on driver fatigue detection system, and lots of achievements have been made. Based on the analysis and comparison of the previous studies, this system uses a method based on computer vision to evaluate the fatigue of drivers by eyes state.In this paper, the study of driver fatigue detection mainly includes four parts:extracting face region, eye detection and tracking, iris location, judging the fatigue.1. Face region extraction. Based on an analysis of existing face detection algorithms, we choose the AdaBoost method based on Haar feature, and it is verified according to the size of face in image. The experimental result has shown that this method could meet the study requirements in detection time and detection accuracy.2. Eye detection. In this paper, we have improved the histogram equalization method and proposed weighting approach to deal with the human eye area. We have proposed the method that using the histogram to select threshold and converted the eyes region to binary image, which has provided a good foundation for the segmentation of eyes region. We have proposed an eye detection method based on contour extraction. This method can remove the eyebrows, nose shadow, the side shadow and other interferences.3. Iris location. We have optimized the existing algorithm, and it can detect the location of the iris accurately.4. We have used the template matching methods to track eyes. After analyzing the factors that affect the running time and then taking the measures of reducing resolution ratio to increase the tracking efficiency. Seen from the experiments, the methods of reducing the resolution ratio do not reduce the tracking precision, but time is greatly reduced.5. Judging the fatigue. Calculating the PERCLOS value with the blink frequency and eyelids distance which has been Computed through the iris, then discussing the calculation of PERCLOS under the condition of no eyes detection. In this paper, we also match the tracking eyes and judging fatigue together, and greatly reduce the detection time of the whole system.