Driver Fatigue Detection Based On Feature Corner Tracking

Traffic accidents caused by driving fatigue have become one of the main factors in traffic safety. Therefore, how to detect and prevent fatigue driving effectively is of profound practical significance, for it can dedicate to reduce traffic accidents and casualties.In the paper, a method of driving fatigue detection based on feature corners' tracking is proposed, which includes face detection, eye corners detection, corners based tracking and fatigue estimation. Before eye corners detection is implemented, face detection is handled firstly to reduce the region and speed up eyes detection. In a series of steps, the requirements of real-time and accuracy are fully taken into account, for instance, only the necessary components in the color space are converted, and the aspect ratios of all the color regions is computed to determine the real face region. Then, during the process of eye corners detection, the method called Harris is used to detect eyes' corners. Before corners detection, some useful operations are conducted to reduce the number of pixels, that is, the pixels that are impossible to be detected as corners are removed. After the process of corners detection, corners based tracking using particle filter is proposed. Corners based tracking can achieve good performance, meanwhile, it can work more efficient than eyes based tracking. Finally, the results of tracking are used to analyze and estimate the driver's fatigue state. The combination of perclos and computing pixels of eye regions dedicates to elicit the state of driver fatigue, which can provide a basis for fatigue early warning.Result shows that, driving fatigue detection based on feature corners' tracking can achieve detection robustly.