首页 | 杂志简介 | 征稿简则 | 投稿指南 | 常见问题 | 名词解释 | 刊物订阅 | 联系我们 | English
主动毫米波成像性别识别算法研究
投稿时间:2018-07-05  修订日期:2018-07-20  点此下载全文
引用本文:周健.主动毫米波成像性别识别算法研究[J].红外,2018,39(9):34~40
摘要点击次数: 76
全文下载次数: 60
作者单位E-mail
周健 中国科学院上海微系统与信息技术研究所 zjian@mail.sim.ac.cn 
基金项目:国家自然基金项目(61671439);国家自然基金重点项目(代码F0120,科学部编号:6173000272);中国科学院院地合作项目(YDJDBNJ-2012-027);上海市自然科学基金(13ZR1447400)
中文摘要:针对主动毫米波成像安检系统的个人隐私保护问题,提出了一种主动毫米波人体图像性别识别流程,包括图像预处理、人体躯干图像分割、性别特征提取、各特征值权重确定、性别识别图谱形成、基于图形数据库的图像分类识别和算法优化验证。提出了一种基于灰度值提取的局域直方图性别识别特征图谱,给出了不同图像主体直观、量化的性别特征表达。基于该图谱,提出了一种多特征、多权值性别识别方法,性别识别准确率高于80%,可对海量人体图像快速分类与筛选,从而采取针对性隐私保护措施。该算法已应用于毫米波安检产品,其可移植性高,运行速度快,识别率高,在同类安检设备中具有巨大的应用潜力。
中文关键词:毫米波  性别识别  图像处理
 
Study of a Gender Identification Algorithm for Active Millimeter-wave Imaging
Abstract:A human body image gender identification algorithm was proposed for active millimeter-wave imaging security systems. The procedure of the algorithm includes imaging preprocessing, human body image segmentation, gender feature extraction, selection of weight of different characteristics, formation of gender feature maps, image classification identification and algorithm optimization verification based on map database. A kind of local histogram algorithm based on gray value extraction was proposed and an intuitive map for identifying gender of the figure was given. On the basis of the map, a multi-weight, multi-characteristic gender identification method was proposed. The identification rate of the method is above 80%, which would be helpful to the fast identification of huge number of pictures and the taking of corresponding measures to protect human privacy. The algorithm had been used in millimeter wave body security checkpoints. Because of high portability, fast running speed and high recognition rate, it would have great application potential in similar security equipment.
keywords:millimeter-wave  gender identification  image processing
查看全文  查看/发表评论  下载PDF阅读器

版权所有:《红外》编辑部

北京勤云科技发展有限公司