Research And Implementation Of Driver Fatigue Detection

Nowadays, with the improvement of living standards of the masses, there are more and more people holding cars. However, traffic accidents are also increasing along with the improvement. Many studies show that driver’s fatigue is an important cause for the growing of traffic accidents. Therefore, it is of great significance to carry out driver fatigue detection and pre-warning research.In reference to a large number of domestic and foreign literatures, this thesis analyzes the characteristics of the actual driving conditions and the requirements of the system, then finds the appropriate way to implement the driver fatigue detection system which consists of face detection module, eye location module, eye states detection module, eye tracking and fatigue analysis module. In the face detection module, a mature face detection algorithm is adopted, that is Adaboost cascade classifier algorithm based on the characteristics of Haar, and based on this algorithm locating the position of face. Then the image preprocessing technologies is used to eliminate the illumination effects of the located facial images. The eye location module use gray integral projection algorithm and block complexity algorithm to locate the eye, namely according to the horizontal projection of the eye to locate the approximate location of human eyes, and then analyzes the complexity of eyes area to locate the exact location of eyes. The eyes state recognition module is according to the human eyes information to determine whether the eye is in a state of fatigue by the standard given beforehand, so as to achieve the aim of fatigue detection. In the eye tracking and fatigue analysis module, the thesis adopted the method combined the PERCLOSE and the blink rate to recognize the state of eyes which enhance the reliability of the system and reduce the traffic accidents.This thesis implements the system in the platform of visual studio; then tests the system with the face database and homemade face library. Experimental results show that these methods can be used to recognize eye state and detect fatigue in the condition of head tilt, light changes, expression changes with high accuracy and good real-time performance. Eyes locating accuracy rate is94.67percent, state recognition accuracy rate is98%, the average processing time for each frame is about50ms. At the end, the insufficiency of the system and the point need to improve are analyzed.