摘要
介绍行人重识别技术的基础理论,并对智慧安防、智能寻人领域常用的数据集和评价指标进行了总结。为满足工程需求,在数据集上对一种基于全局特征的行人重识别网络模型进行了训练和测试。对基于行人重识别的目标跨镜追踪系统进行设计,使系统具有行人目标图像匹配和行人目标跨镜追踪两大核心功能,以便有效地对目标人员进行多场景下多摄像头的持续追踪。
This paper introduces the theoretical basis related to person re-identification and summarize and outline the commonly used datasets and evaluation in the field of intelligent safety and intelligent people finding.For engineering requirement,in the datasets,a person re-identification network model based on global features is trained and tested.Based on person re-identification,a pedestrian tracking system is designed,achieving two core functions of pedestrian target picture matching and pedestrian target cross-camera tracking,which can effectively perform continuous multi-camera tracking of target persons in multiple scenes.
作者
韩珂
贾冕
王乾隆
HAN Ke;JIA Mian;WANG Qianlong(College of Information Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450046,China)
出处
《成组技术与生产现代化》
2022年第3期21-27,共7页
Group Technology & Production Modernization
基金
河南省高等学校重点科研计划资助项目(20B520014)
河南省高等教育教学改革研究与实践重点计划资助项目(2021SJGLX167)。
关键词
深度学习
行人重识别
计算机视觉
跨镜追踪
追踪系统
deep learning
person re-identification
computer vision
cross-camera track
tracking system
