基于DSP的嵌入式驾驶疲劳监测系统研究

Research on Embedded Driving Fatigue Monitoring System Based on DSP

为了进行有效的、实时的、非接触式的驾驶员疲劳检测,采取NHTSA和FHWA推荐的非接触式疲劳检测方法PERCLOS,利用TI公司的专用图像处理DSP芯片TMS320DM642,构建了车载的基于DSP的嵌入式驾驶疲劳监测系统。首先使用肤色模型算法迅速定位人脸,然后利用累积差分帧和Hough变换等实时图像处理技术来检测、跟踪眼睛,分析眼睛的状态和提取眼睛的特征参数,计算PERCLOS值、判断疲劳程度并采取报警或其他措施。试验结果表明,系统的算法简单、快速、鲁棒性强。

To do effective, real-time, non-contact detection of driving fatigue, PERCLOS is a non-contact method that is recommended by NHTSA and FHWA, and carbome embedded fatigue driving monitoring system is constructed using TMS320DM642. At first human face is localized using skin color model, then ADF and Hough transform technology of the real-time images are used to detect, track and analyze the state of eyes and pick-up characteristic parameter of eyes, count closing time of eyes in succession within regular hour, judge the tired degree and take the corresponding measure in value of calculating out PERCLOS.Experiments show that the algorithm is simple, fast and robust.