SLAI Seminar 21st Session will be discussing the topic on "New Advances in Person Re-identification——From Individual Pedestrians to Small Groups, From Ground to Air-Ground Integration ", from 4pm to 5:30pm, January 12th (Monday) at B411 Lecture Hall, online participation is welcome (Tencent Meeting ID: 840-739-186)

报告主题：行人重识别新进展——自单一行人到小小群体，自地面到空天一体化

时间：2026年1月12日（周一）16:00-17:30

地点: 深圳河套学院B411阶梯教室

线上参与：腾讯会议号840-739-186

讲者简介 About the Speaker:

赖剑煌教授，二级教授，博士生导师，现任职于中山大学计算机学院。中国图象图形学学会会士、名誉副理事长；广东省图象图形学学会第四届、第五届理事长。中国计算机学会杰出会员；中国计算机学会计算机视觉专委会第一、二届副主任；广东省人工智能与机器人学会第一届副会长；广东省公共安全技术防范协会人工智能专业委主任。赖教授于1986年、1989年分别获得中山大学学士、硕士学位并留校任教，1999年获中山大学博士学位。主要研究方向包括计算机视觉、模式识别、机器学习等。赖教授主持国家自然科学基金-广东省联合重点项目、科技部科技支撑计划、国家自然科学基金等项目，曾获广东省科学技术奖自然科学奖一等奖（2018，排名第一）、丁颖科技奖（2019），并享受国务院政府特殊津贴，现已发表学术论文约200篇，主要刊登于ICCV、CVPR、ICDM等国际顶级会议，以及IEEE TPAMI、IEEE TIP、IEEE TNN、IEEE KDE、Pattern Recognition等国际权威期刊。

Lai Jianhuang Professor (Second-Class), PhD Supervisor, School of Computer Science, Sun Yat-sen University. Honorary Vice Chairman and Fellow of the Chinese Society of Image and Graphics (CSIG); Chairman of the Guangdong Society of Image and Graphics (4th & 5th Terms). Distinguished Member of the China Computer Federation (CCF); Deputy Director of the CCF Technical Committee on Computer Vision (1st & 2nd Terms); Vice Chairman of the Guangdong Artificial Intelligence and Robotics Society (1st Term); Director of the Artificial Intelligence Professional Committee of the Guangdong Security and Protection Association.He received his Bachelor’s and Master’s degrees from Sun Yat-sen University in 1986 and 1989 respectively, and has been teaching at the university ever since. He obtained his PhD degree from Sun Yat-sen University in 1999. His main research fields include computer vision, pattern recognition, and machine learning.He has presided over a series of projects, including the Joint Key Project of the National Natural Science Foundation of China and Guangdong Province, the Science and Technology Support Program of the Ministry of Science and Technology, and projects funded by the National Natural Science Foundation of China. His academic achievements have been recognized with numerous awards, such as the First Prize of Natural Science Award of Guangdong Provincial Science and Technology Award (2018, Rank 1), the Dingying Award 丁颖奖 (2019), and he enjoys the Special Government Allowance of the State Council 国务院政府津贴.He has published approximately 200 academic papers, mainly in top-tier international conferences including ICCV, CVPR and ICDM, as well as authoritative international journals such as IEEE TPAMI, IEEE TIP, IEEE TNN, IEEE KDE and Pattern Recognition.

报告摘要 Abstract：

小群体重识别旨在精准关联非重叠监控视野下具有相同成员组成的群体图像。作为传统行人重识别任务的重要延伸，小群体重识别在安防监控场景中具有重要研究价值与广阔应用前景。其独特挑战在于建模群体成员数量与空间布局的变动性，提取稳定鲁棒的群体特征表达。近年来小群体重识别受到研究者广泛关注并取得快速发展。本次报告主要介绍小群体重识别技术的科学问题及相关研究进展，包括我们实验室的研究探索与实践成果：基于显著性关键点、孪生网络、不确定性建模与群体三维布局重构的跨视角鲁棒群体特征提取方法；基于单人-群体距离和最近排列距离的群体度量方法；基于行人-群体关联性的跨域重识别技术；以及City1M等大规模虚拟现实群体数据集与CMGroup等跨模态群体数据集的构建。

Small group re-identification aims to accurately associate group images with identical members captured by camera networks under non-overlapping fields of view. As a crucial extension of traditional person re-identification tasks, small group re-identification holds significant research value and promising application prospects in security surveillance scenarios. The unique challenges of small group re-identification lie in modeling the variations in the number and spatial layout of group members, as well as extracting stable and robust feature representations. In recent years, small group re-identification has attracted extensive attention from researchers and achieved rapid development. This report mainly introduces the scientific issues and related research progress of small group re-identification technology, including the research exploration and practical achievements of our laboratory: robust group feature extraction methods in cross-view scenarios based on salient key points, siamese networks, uncertainty modeling, and group 3D layout reconstruction; group metric methods based on person-group distance and nearest permutation distance; cross-domain re-identification technology based on person-group correlation; and the construction of large-scale virtual reality group datasets such as City1M and cross-modal group datasets such as CMGroup.