基于加权特征融合与局部特征注意的人种分类  

Race classification based on weighted feature fusion and local feature attention

作　　者：董永峰 钟璨[1,2,3] 齐巧玲 李林昊 DONG Yong-feng;ZHONG Can;QI Qiao-ling;LI Lin-hao(School of Artificial Intelligence,Hebei University of Technology,Tianjin 300401,China;Hebei Province Key Laboratory of Big Data Computing,Hebei University of Technology,Tianjin 300401,China;Hebei Engineering Research Center of Data-Driven Industrial Intelligent,Hebei University of Technology,Tianjin 300401,China)

机构地区：[1]河北工业大学人工智能与数据科学学院,天津300401 [2]河北工业大学河北省大数据计算重点实验室,天津300401 [3]河北工业大学河北省数据驱动工业智能工程研究中心,天津300401

出　　处：《计算机工程与设计》2024年第9期2683-2689,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61902106);河北省自然科学基金项目(F2020202028)。

摘　　要：为充分利用浅层特征中的细节纹理信息对人种特性的描述能力,挖掘具有区分性部位的表达特征对人种分类的作用,更好利用数据不同层次的特征与区分性部位以提供更具鲁棒性的人种信息,提出一种基于加权特征融合与局部特征注意的人种分类模型(weighted feature fusion and local feature attention model,WFLA)。模型设计加权特征融合模块增强浅层与深层特征的交互,构建局部特征注意模块重点关注区分性部位。在3个公开数据集中的大规模验证实验验证了WFLA模型在人种分类任务中具有明显优势。To make full use of the description ability of the detailed texture information in the shallow features on the race characteristics,explore the function of the expression features of the distinguishing parts on the race classification,and better use the features and distinguishing parts at different levels of the data to provide more robust race information,a population classification model based on weighted feature fusion and local feature attention model(WFLA)was proposed.A weighted feature fusion module was designed to enhance the interaction between shallow and deep features,and a local feature attention module was constructed to focus on distinguishing parts.The large-scale verification experiments in three public data sets demonstrate that the WFLA model has obvious advantages in the task of racial classification.

关 键 词：人种分类 注意力机制 多层融合 深度学习 局部特征 特征提取 特征交互 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]