The Research Of Drivers' Tired Inspection Based On Face Recognization

Along with the development of the economy, the living condition is rising and the traffic condition is developing. According to the relative reports, the tired driving is the third in the dangerous factor which cause traffic accident. In the recent years, many of the serious accidents were relate to the tired driving directly or indirectly. So how to know if the driver is in danger fast and accurately is a common case arouse worldwide attention. Many automobile manufactory in different countries tend to solve the problem .But now there is no mature skill to copy with it. As the image measuring technique has many advantages such as non-contact, the high velocity, the rich information content, the low cost and so on. So we can apply this technique to the area of the tired identification, and it must be a challenge subject .So I support the tired identification base on the face recognition, and it just a little research in the area of tire identification .Following is the main parts of my paper:Chapter One: Analyses and compare the various ways to solve the problem .After that I lifted some typical measurements and compare their own advantages or disadvantages. I also point out the major parts of my research and its important parts. The task of my research is to set up a real-time processing system of the image, use which to measure the drivers'performance while he is driving. The real-time of the measuring system is difficulty depart because the processing time must be less than 100 milliseconds and beside it we must take the effect taken by the change of the sunlight into consideration.Chapter Two: we take the effect taken by the change of the sunlight into consideration. So we apply median filer and histogram equalization to image enhancement. And in order to avoid the serious effect of gray taken by the change of the sunlight, we apply the edge of face to the image process, and also it can help us to reduce the information content of the image. In the respect of image edgedetection, I carry on the comparison with all kinds of tradition algorithms(Roberts,Sobel,Prewitt,Laplacian), elaborates the principle of each algorithm. After that I choose the best algorithm in the respect of the detection effect and the real-time.Chapter Three: This paper analyses the key element of matching algorithm namely matching primitive, measurement function, search strategy, evaluation mechanism and in common use matching algorithm, and presents their applicable instances and their relative merits.Chapter Four: We discuss and analyses the various ways in face detection, including face detection based on the characteristic and face detection base on the image. Elaborate the face matching algorithm based on the edge of image and then show the fast face image matching based on grid division and its vector. We use the reduced eight directions vector to describe face and establish the digital matrix of the face. So we can use it as a matching templet. From the absolute error among the matrix, by computing the similarities and shape alteration between the vector, the course of the goal matching recognition is accomplished. After that I analyses the major factors which lead to the mismatching and how to correct it. Enumerate some methods to recognize the eyes, including gray projection,using the templet which is composed by eyes and nose,nerve net, and presents their applicable instances and their relative merits. First of all we should set up a templet which is composed by eye and eyebrow and use it to find the similar area as the templet. After that we make horizontal and vertical projection of the area which is found by using the templet. So we can know the coordinate of the eye. We can know the effect of this algorithm and what affect the algorithm by test. Finally I explain how to use the result to identify if the driver is in danger briefly by whether the fettle of the drivers'eyes have changed in a period of time.Chapter Five: Some researches for hardware implementation of target identification system are made in this paper. The identification system should be small, easy in debugging , all-purpose and so on. Image sensor7131 is adopted in image collecting, the time of gathering a frame image is 21ms. Storage chip isIC61LV25616, Thereof storage capacity is 256K16bit. FPGA accomplishes image data collection, memory, template region matching and communication with ARM, Cycloneâ…¡EP2C5 of Altera company is adopted. Identification algorithm is achieved by ARM, the chip is LPC2106. Arm takes over the results of region template matching from FPGA, and analyses statistics the fettle of the drivers'eyes...