Research On The Method Of Eye Detection And Tracking Based On DSP

Driver fatigue is one of the most important reasons of traffic accident. It is of great social significance to reduce the proportion of traffic accidents through technical means to detect the driver fatigue. At present, Scholars from many countries have carried out many ways to detect the driver fatigue. Real-time, vehicle-mounted and non-contact detection means are the hot topics of the present study, of which the method based on PERCLOS principle is the most popular way. The proportion of the closing time is detected and calculated to judge whether driver is in fatigue condition. The emphasis and difficulty of this method is to track the driver’s eyes in real time accurately.The simulation study of the algorithms in this paper is carried out in Matlab, and the experiment is tested on the SEED-VPM642. The main contents in this paper include illumination processing, the driver’s face tracking technology, as well as eyes tracking technology research in real-time.The main research work in this paper is as follow:(1) The detailed analysis of research on driver fatigue detection technology from both objective and subjective testing technologies is carried out, and the characteristics of the various detection methods are summarized and compared in this paper;(2) The study platform for driver fatigue detection system based on DM642is built in this paper. In this system, the SONY CCD camera is used as the input device, the A/D module TVP5150AM1is used to convert the analog signal into digital signal, then the signal is saved in the video storage module, then the signal is processed in the TMS320DM642through the algorithm studied in this paper, and then the module SAA7172H module converts the digital signal into analog signal, and last the video is displayed in the display. The development board is connected to the computer through SEED-XDS510PLUS, and the software is debugged through CCS3.3which is installed on the computer.(3) The light preprocessing technologies on the video image are summarized and researched in this paper. First, the video image is processed through a histogram equalization program to enhance the contrast of the video. Through automatic detection and correction techniques of highlight area, the pixels which are changed because of the light are corrected; thereby the accuracy rate of the system is improved.(4) The driver face tracking technologies based on DSP are summarized and researched in this paper. Based on the light preprocessing technology, the similarity calculation is carried out in YCbCr color space, and the pixel value is normalized, and then the Otsu algorithm is applied to image binarization, and then the video is smoothed through corrosion expansion technology, al last, the face boundary is determined by the calculation of horizontal and vertical integral projection of the images.(5) The eye detection and tracking technologies based on DSP are summarized and researched in this paper. First, the probable eye region is detected by the color gray integral projection, and the accurate eye region is detected according to the positional relationship of the eyes and eyebrows, then the Kalman filter is used to track the eyes in real-time. After a certain time, in the process of eye tracking with Kalman filter, considerable error will happen, so it has to reposition the area of the eyes.