摘要
针对传统接触式签到依赖服务器等问题,提出一种基于轻量化人工智能人脸识别算法,设计嵌入式终端无接触式签到系统。利用预训练的MobileFaceNet模型提取人脸特征,将其输入Siamese模型得到降维后的特征向量,计算该向量与特征库中向量的最小欧式距离,并与阈值对比完成人脸识别。系统以STM32MP1微处理器为CPU,利用OpenCV和TensorFlow Lite软件实现签到功能。结果表明,系统具备良好的实时性和稳定性,可通过模型在线更新添加新人特征,提升人脸识别的准确率和实用性,符合当前疫情防控政策下的实验室签到需求。
In order to solve the problems of traditional sign-in system relying on servers,a face recognition algorithm based on lightweight artificial intelligence is proposed,and a contactless sign-in system based on embedded terminals is designed.The face features are extracted by pre-trained MobileFaceNet model,subsequently are input into the Siamese model to obtain reduced dimensional feature vectors,and then the minimum Euclidean distance between the feature vectors is calculated.Finally,the face recognition is achieved by comparing the minimum Euclidean distance with the preset threshold.This system uses STM32MP1 microprocessor as CPU,and is achieved by using software such as OpenCV and TensorFlow Lite.The results show that this system has good real-time performance and stability.The face features of new people can be added via online updating of the Siamese model,thus it can improve the accuracy and practicability of face recognition.The system meets the laboratory sign-in requirements under the current epidemic prevention and control policy.
作者
陆玲霞
强柱成
于淼
任沁源
LU Lingxia;QIANG Zhucheng;YU Miaoa;REN Qinyuan(College of Electrical Engineering,Zhejiang University,Hangzhou 310027,China;College of Control Science and Engineering,Zhejiang University,Hangzhou 310027,China)
出处
《实验室研究与探索》
CAS
北大核心
2023年第3期130-134,170,共6页
Research and Exploration In Laboratory
基金
国家自然科学基金联合基金项目(U21A20485)
浙江省基础公益研究计划项目(LGG22F030008)
浙江省高校实验室工作研究项目(ZD202103)
教育部产学合作协同育人项目(202002302017)
浙江省产学合作协同育人项目(202018)。
