从虹膜图像中揭示人类性别与种族研究综述OA

This paper systematically reviews the research progress in demographic analysis based on iris images,with a focus on the classification of two key attributes:gender and race.It elaborates on the fundamental principles and task diffi-culties of iris attribute classification,followed by an introduction to its application scenarios and two integration modes with iris recognition(serial integration and parallel integration).Subsequently,seven commonly used quantitative evalua-tion metrics for iris attribute classification are listed.For iris gender classification,the developmental trajectory is presented through two major categories:traditional handcrafted feature-based methods and deep learning-based methods.For iris race classification,the progress is outlined via statistical value-based methods and texture primitive-based methods.Addi-tionally,the paper summarizes the characteristics and applicable scenarios of mainstream iris attribute classification datasets,including ND-GFI,CVBL_IRIS,UNAB-Gender,and CASIA-BioSecure.Furthermore,it analyzes current challenges in iris attribute classification,such as attribute classification in unconstrained environments,multimodal/feature fusion,cross-domain(spectral/sensor)generalization,data distribution and annotation,and model lightweighting.Finally,the paper concludes with a comprehensive summary,providing systematic references for technological development and applica-tions in this field.