论文部分内容阅读
疲劳驾驶是引发恶性交通事故的重要原因之一,驾驶员疲劳监测技术近年来已逐步成为图像处理领域的一个研究热点。基于改进的和提出的新算法,设计了一个嵌入式驾驶员疲劳监测系统。由可见光/近红外摄像头采集视频,首先采用Haar特征的级联分类器从图像中检测出人脸区域,并用钻石搜索法跟踪人脸区域;然后提取一个新的图像差分统计特征,并结合3个准则判断疲劳状态;最后采用全变分模型消除图像中的非均匀光照,以便实现鲁棒的人眼定位和人脸识别。实验测试结果表明,本系统的疲劳状态监测准确率达到95%以上。
Fatigue driving is one of the important causes of the vicious traffic accidents. Driver fatigue monitoring technology has gradually become a research focus in the field of image processing in recent years. Based on the improved and proposed new algorithm, an embedded driver fatigue monitoring system is designed. The video is captured by a visible / near-infrared camera. First, the face region is detected from the image using the cascade classifier with Haar feature and the face area is tracked by the diamond search method. Then, a new differential statistical feature of the image is extracted and combined with 3 Criteria to judge the state of fatigue; Finally, the use of total variation model to eliminate non-uniform illumination in the image, in order to achieve robust human eye location and face recognition. Experimental results show that the fatigue monitoring accuracy of this system reaches more than 95%.