可视化智能门锁中的3D人脸识别性能测试方法的探究  

Research on the Performance Testing Methods of Facial Recognition in Visual Smart Locks

作　　者：郑益衡 廖亮 张传甲 刘懿敏 林奔 

机构地区：[1]威凯检测技术有限公司温州分公司,温州325000 [2]威凯检测技术有限公司,广州510663

出　　处：《日用电器》2024年第11期112-116,130,共6页ELECTRICAL APPLIANCES

摘　　要：面对3D人脸识别算法在可视化智能门锁上广泛应用的现状,本文通过对比目前智能门锁市场上常用的3D人脸识别技术,提出了一种以认假率(FAR)和拒真率(FRR)为核心评价指标的3D人脸识别性能试验方法。本文详细介绍了该方法的测试人脸库的选择、测试流程的设计和评价指标的统计方法,并通过实际案例的试验,为智能门锁人脸识别技术的性能测试方法提供了一种参考。In light of the widespread application of 3D facial recognition algorithms in visual-based intelligent locks,this paper proposes a performance testing method for 3D facial recognition,focusing on two core evaluation metrics:False Acceptance Rate(FAR)and False Rejection Rate(FRR).By comparing the commonly used 3D facial recognition technologies in the current intelligent lock market,the paper provides a detailed explanation of the selection of the fa-cial database for testing,the design of the testing process,and the statistical methods for the evaluation metrics.Through practical case experiments,this paper offers a reference for the performance testing methods of facial recognition tech-nology in intelligent locks.

关 键 词：3D人脸识别 认假率 拒真率 人脸库 特征项 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]