Research On Fatigue Driving Alarm System Based On Facial Features

Since the late of 1970s,people’s lives were more and more to be good. In the past, cars were too expensive to many Chinese. But now, it was not. As a modern ways of transport, more and more Chinese ordinary people enjoyed them. By the end of 2014, we had locomotives with 264millions,of which the number of cars with 154 millions. And people’s traveling was more and more convenient. However, cars were not always giving the benefit to the people, they also bought disadvantages to humans. Accidents happened almost every day of our lives. Some people might be injured, others might lost their lives. Traffic accidents had become a social problem. The modern fast-paced city life, more and more people were inability to concentrate on something for a long time, especially after a long time driving. They were often tired for physical and mental, and many traffic accidents occurred in such situation. Especially in some of the major accidents, accidents caused by fatigue driving was occupied a large ratio. The accidents bought a huge number of problems with many people’s lives and the social stability. Therefore, developing an effective early warning system of fatigue driving to reduce traffic accidents and improve drivers safety was imminent.After analyzing the present situation of researching fatigue driving, the paper analyzed the person’s face organ distribution location and their status. We obtained satisfactory results after doing the simulation tests on the software MATLAB. In this paper, the next researches were done:(1)Firstly, we analyzed the background and significance of the paper according to the many negative factors caused by fatigue driving in our country. Secondly, according to the domestic and the abroad analyzing with this subject based on extensive literature knew the current research status. Finally, we discussed the direction of research and development about fatigue driving warning system.(2)According to the characteristics of skin color, we compared several common color space model. As was known to all, the skin color was different, but the reason is the brightness. Therefore, the face was selected based on YCbCr color space and established the Gaussian model. Since Adaboost algorithm could filter out small portions that might interfere to identify face, so we also studied the Adaboost algorithm. The template matching method was based on the face gray average. We must be prepared the template pictures in advance, and then calculated the correlation,the correlation was more smaller the result was more good. In this paper, in order to get more accurately face picture and get ready to extract features in the follow thesis, we integrated these three methods. The experimental results showed us that the speed was more quickly and the accuracy was more good.(3)Then this thesis analyzed and studied the characteristics of the eyes and mouth, and to detect the eyes and mouth and position them. At the first time, we analyzed the common positioned methods; at second, in order to consider the speed and the efforts made in the last, the improved "three court five" method was proposed. We could position the eyes and mouth according it. And then, the picture was converted to the binary image. Lastly, we could extract the eyes and mouth accurately according the contours of the eyes and mouth.(4)After correctly positioning of the eyes and mouth, the next step was to analyze their state of fatigue or non-fatigue. For the eyes, we use the classic judgment eyestrain algorithm, PERCLOS algorithm; for the mouth, the article is based on the template matching method, that we prepared the template images that which mouths were open and which mouths were closed. And last, the getting images could match to the template images.(5)Whether it was based on the eyes’state or mouth’s state, it was just based on a single factor. It was very different to accurately determine the drivers were tired or not. It was very funny, information fusion technology was to make up for this deficiency. After analyzing several common research information fusion algorithms, we took DS evidence theory inference algorithm. We fused the information of the eyes’state and mouth’s state by DS algorithm. In the end, we determined the drivers who were fatigue or non-fatigue driving by considering two factors.This article fully considered the face located and facial features(eyes and mouth) extracted. Then, it was analyzing the states of eyes and mouth. The next step, we could fused the eyes’ state information and the mouse’s state information by DS algorithm. The finally, we could judge the drivers who was fatigue or non-fatigue.