| Fatigue driving is an important factor in causing traffic accidents. Therefore, it is of great means to design an all-weather system which can monitor drivers' fatigue degree in real-time to avoid the accidents and guarantee personal safety. This paper studies how to realize the all-weather system based on DSP by using the technology of computer vision. The main contents of the paper are described as follows:1. An all-weather system which can monitor drivers' fatigue degree in real-time is built based on DSP (TMS320DM643). An infrared night-vision CCD camera is adopted as the system's input device. Colored images are got when the light is abundant, and infrared images are got while the light is gloomy.2. The algorithm of face location is studied. As for colored image, a method of skin-color model combined with regional area is proposed which can quickly locate the face in color images. As for the infrared images' face location, the dynamic background update algorithm and the discontinuous frame difference algorithm are studied and compared under the experiment conditions.3. The algorithm of eye location is studied. A method combining five eyes of three courts with vertical integral projection is proposed to divide the right eye from the left eye, the target eye which will be selected to be located in next step is chosen between the left eye and the right eye by area. As for colored images, the difference projection function and the integral projection function are used to locate the target eye. As for infrared images, a method combining the difference projection function with the reflective properties of the eye under near IR illumination is proposed to get the target eye's two possible eye candidates. In addition, the target eye's state is estimated straightly on the two possible eye candidates. The experiment results indicate that when the target eye is located, the proposed scheme can avoid interference from eyebrows and nose, etc., meanwhile, the system's computational complexity is reduced owing to overleap the target eye's precise location.4. The eye's state is estimated by the ratio of eye's height and eyeball's diameter, the method of fuzzy decision is used to estimate the drivers' fatigue degree based on the state of the target eye, which integrates the parameters of PERCLOS, ECT and EBR The method can reduce fatigue state estimation's uncertainty made by the single parameter.5. The algorithms' debugging is completed on DSP (TMS320DM643). These algorithms concretely include face location, eye location, eye state estimation, fatigue parameters extraction and fatigue state estimation. The experiment results show that the system can accurately monitor the drivers' fatigue degree in real-time. |
