摘要
现有的识别亲缘关系的方法大多数仅能识别单亲亲缘关系(父子关系、父女关系、母子关系、母女关系),并且这些识别方法对年龄差距大、性别不同的父女或母子等识别样本的识别效果不佳.为了解决这些问题,提出一种可以同时识别孩子与父母之间关系(双亲亲缘关系)的识别方法.鉴于目前针对双亲亲缘关系的识别方法极少,提出一种基于度量学习和相关分析的识别双亲亲缘关系模型来提高识别子女与双亲之间亲缘关系的准确度.依据子女与双亲的生物遗传关系设计可融合子女与双亲的亲缘特征的多线性并行网络;并利用判别式度量学习和典型相关性分析在数据处理中优势,从包含多种人体身份的面部特征中提取有利于亲缘关系识别的特征信息,用于鉴别子女与父母是否存在血缘关系以实现识别精度的提高.实验结果表明,所提出的方法在子女与双亲的亲缘关系识别上效果更好.
The existing kinship recognition methods can only recognize single-parent relationships(eg.father-son,father-daughter,mother-son,mother-daughter).However,these methods have certain limitations and are not effective in recognizing samples of parents with large age gaps and different genders.In order to solve the above problems,this paper proposes a method of identifying kinship(based on parental relationship).As there are few methods for recognizing the parental relationship,this paper proposes a kinship model for recognizing parental relationship based on metric learning and correlation analysis to improve the accuracy of recognizing kinship.First,based on the DNA of children and parents,a multi-linear parallel network of integrating the kinship characteristics of children and parents is designed.Then,using the advantages of discriminative metric learning and canonical correlation analysis in data processing,the facial features benefitting kinship recognition are extracted from multiple human identities,which are used to identify whether there is a relationship between the children and their parents to improve the accuracy.Experimental results show that the proposed method has a better effect on the recognition of the kinship between children and there’s parents.
作者
孙劲光
贾彦勇
宋晟民
SUN Jin-guang;JIA Yan-yong;SONG Sheng-min(College of Electronic and Information Engineering,Liaoning Technical University,Huludao 125105,China)
出处
《控制与决策》
EI
CSCD
北大核心
2021年第8期1977-1983,共7页
Control and Decision
基金
国家自然科学基金项目(61702241)。
关键词
亲缘关系识别
度量学习
相关性分析
迁移学习
深度学习
数据扩充
kinship relationship recognition
metric learning
correlation analysis
transfer learning
deep learning
data augmentation
