A Driver Fatigue Detection System Based on Eye States Tracking

• Main Authors: Chia-chuan Chang, 張家銓
• Other Authors: Ming-shi Wang
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/03940611027855608541

碩士 === 國立成功大學 === 工程科學系碩博士班 === 97 === In this thesis, a driver fatigue detection system based on tracking driver’s eye states was implemented. The system contains five parts: face detection, eye position detection, eye tracking, recognize eye state and fatigue detection. Firstly, skin color and projection methods are used to get the face area of an image. In eye position detection, the possible location area of eyes on the face area is selected and then the Sobel vertical operator is used to get the edge image of the selected subimage. Again, projection method is used for the edge image to obtain the eye’s position. Particle filtering method, according to the Gaussian distribution is adopted to perform the eyes tracking of the video frames. Two templates, eyes open and eyes closed are used to estimate the weights of the particles and to decide if the eyes are open or closed. Finally, the rate of eye closed frames within a certain period of time can be used to determine whether the driver fatigues or not. For evaluation the proposed system, 9 videos which taken in the day time with varied brightness were tested. It is shown that the system always gives a correct result.