Driver Fatigue Real-time Monitoring System Design And Realize Based On DSP Technology

The invention and popularization of transportation vehicles such as cars,trains and planes promote the social great development, and also provide people great convenience. But everything has two sides, The emergence of various transportation vehicles brought us a large number of disasters. According to relevant statistics:Every year, the number of deaths caused by traffic accidents up to 60 million, The direct economic losses is about 12.5billion dollars, Fatigue-driving and Drink-driving are the main factors. Once the driver felt fatigue, he could not be concentrated on the sudden decline in the resilience of the situation, physically unresponsive, easily lead to accidents. For these reasons, Driver fatigue warning system should be developed urgently.In this paper, TI'sTMS320DM6437 high-speed image processing chip was used as the main processing unit. The infrared light micro-camera was used to capture driver's face image. By analyzing the fatigue presentation and the fatigue reasons, The driver fatigue was judged by the eye states. When shine by one side or shine not enough, The image which adopt by the CMOS mini camera would be light unevenly. Compare and analyze the Homomorphic filtering method, histogram equalization and gray enhancement, Homomorphic filtering method considered reflection characteristic of the light and high frequency and low frequency of the image. The face skin color has the cluster character so the system completed face detection base on skin color RGB cluster. This method is simple and effective, suited for the simple background face recognition.Eye states was chose as the fatigue characteristic parameters, So eye location and tracking directly impact on the results of fatigue recognition. Edge detection and Binarization was compared, the experiment shows that Ostu method is good. Then eyes located by hybrid integration in the original image based on segmented image feature. Mean-Shift method used to track the eyes. By analyzing eye motion state, Fatigue characteristic parameters (including PERCLOS, eye closure time, eye blind frequency, PERCLOS is now acknowledged as the most effective, vehicle and real-time evaluation indicator of driver fatigue system)were chose. In order to ensure fatigue characteristic parameters were not changed by special states, The average fatigue characteristic parameters in a period time were chose as fatigue characteristic parameters of the system.During the decision-making stage, The fatigue characteristic parameters was fused by information fusion method, and then judge whether the driver is fatigue. If it is true, Alarm module give the driver warning to avoid accidents. The using of information fusion method unproved the accuracy of fatigue monitoring. In the Curse of study follows, if the system add other fatigue parameters, this method can also be used, That would be convenient to improve this system.The system uses dedicated DSP chip developed environment CCS3.3, it should be configured for different DSP model before used. The plugs in CCS3.3 environment such as DSP/BIOS and RTDX play great roles in Online simulation debugging.This driver fatigue monitoring system is sophisticated, convenient, and with high accuracy.