摘要
人脸识别目前应用于多个领域,受控性人脸识别现在有很高的准确度,超过人眼的97.35%,但由于新冠疫情的肆虐,导致大部分人在疫情期间外出都会佩戴口罩,导致识别的准确度降低。因此如何消除口罩遮挡对人脸识别产生的影响,从而提升口罩遮挡人脸识别准确度方面还有很大的研究空间。本文通过提取出未遮挡部分,对未遮挡部分hog特征提取并对提取的特征进行pca降维,通过对嘴巴检测判断是否佩戴口罩,另外加入canny边缘检测作为参考决策。
Face recognition is currently applied in many fields.Controlled face recognition has high accuracy,which is 97.35%higher than that of human eyes.But because of COVID-19’s wreak havoc,most people will wear masks when they go out during the epidemic,leading to the lower accuracy of recognition.Therefore,there is still a lot of research space to eliminate the influence of mask occlusion on face recognition,and improve the accuracy of face recognition.In this paper,the unhindered part is extracted,the features of the unhindered part of the hog are extracted and PCA dimension reduction is performed.The mouth detection is used to judge whether to wear mask or not,and Canny edge detection is added as a reference decision.
作者
侯现坤
周玮
Hou Xiankun;Zhou Wei(Suzhou University,Suzhou Anhui,234000)
出处
《电子测试》
2021年第18期54-55,60,共3页
Electronic Test
基金
2016宿州学院重点科研项目“基于弹性虚拟池的云计算数据中心节能研究(2016yzd10)”
2020年省级大学生创新训练项目“提高口罩遮挡人脸识别准确度的研究(S202010379128)”
2021年国家级创新训练项目“提高口罩遮挡人脸识别准确度的研究(202110379028)”。
关键词
口罩遮挡
HOG特征
区域比例
边缘检测
Mask occlusion
Hog feature
Regional proportion
edge detection
