一种基于面部特征点的疲劳人脸图像检测与识别改进算法研究

An Improved Algorithm for Fatigue Face Image Detection and Recognition Based on Facial Feature Points

研究了基于视频的疲劳人脸检测问题。通过网络爬虫、CEW数据集和现场采集三种方式构建了疲劳人脸检测数据集样本,通过dlib算法识别人面部的特征点,提出了一种基于嘴部开合面积的改进判别策略,作为识别疲劳程度的标志,测试结果表明,算法精度较通用算法提升了8%,达到89%以上,且具有较好的泛化能力,为算法的工程化应用奠定了坚实基础。

Problem of fatigue face detection based on video is studied. Data set of samples of fatigue face are constructed through three methods: web crawler, CEW data set and field collection. Feature points of the human face are recognized through dlib algorithm, and an improved discrimination strategy based on opening and closing area of the mouth is proposed as a recognition method. Test results show that the accuracy of the algorithm reaches 89%, 8% higher than that of the general algorithm, and has good generalization ability, laying solid foundation for engineering application.