A Study Of Driver Fatigue Detection System

With the development of transportation, traffic accidents have become the current serious problems for all countries, and fatigue driving is one of the prime factors that caused a traffic accident. Fatigue detection based on Machine Vision has great advantages in real time, non-contact and all-weather monitoring comparing with other methods, so this method has become a hotspot of current research.On the basis of previous work, this paper gives an effective fatigue detection algorithm. It can be divided into four processes: human face detection and location, human face adjustment, human eyes detection and location, human eyes track and location and identify the state of pilot fatigue. In this paper, the content and main results are as follows:(1) Human face detection and location. Obtain an infrared image from the infrared camera of the video stream, and then make the use of the advantage of that infrared human face region has higher brightness, dark background and easy, and binarize the image with iterative thresholding algorithm, and then locate the face region with the method of region marked. The face detection method can not only locate the face accurately, but also basically free from the effects of light. It resolves the disadvantages of the tradition detection which can not locate the face accurately because of the effects of lights and face pose.(2) Human face pose adjustment. The human eyes detection and location adopts the template matching method, and the eye templates are horizontal and not rotating, then if human face rotates a certain angle, it will result in failing to match, and then get a false template-matching result. Therefore, the human face to be detected must be adjusted by rotation before the process of human eyes detection and location, and this can improve the success rate of human eyes template matching.(3) Human eyes detection and location. Based on detailed analysis of the characteristics of infrared images, this paper improved the traditional template matching algorithm, and conclude an effect template matching algorithm named "multi-steps template matching ". It greatly reduces the number of template matching, and reduces the computation of the algorithm. Because of the samples in the process of the template matching can equably cover the search subgraph, this method ensures the accuracy of template matching. So this method has the advantages with accurate location, high speed, and can satisfies the request of real time.(4) Adopting driver fatigue recognition algorithm, the algorithm of this paper can automatically measure and continuously track the state of pilots' eyes whether opened or not, and according corresponding threshold to remind the driver. Because of the night is the peak time of fatigue driving, this method adopt infrared light as the light source, and it can resolve the problems of the instability of the night light. This method has the advantages with easy and effect, lower complication, and has better fault-tolerance and robustness.