自动化学院“六朝松∙智控论坛”—名家讲坛系列报告

发布者:赵剑锋发布时间:2024-10-17浏览次数:10

报告时间:20241018日 周五上午 9:30

报告地点:东南大学四牌楼校区中心楼二楼中庭

组织单位:东南大学 自动化学院


报告主题:Autonomous Localization Under Low-Cost Multi-Sensors

报告人简介:

Chi Man VONG 教授Chi Man VONG received his PhD from the University of Macau in 2005. He is currently  an associate professor and the associate head of the department of computer and  information sciences in the University of Macau. His current research interests are  computer vision in mobile robotics and autonomous vehicles, and weakly supervised  semantic segmentation. He has published over 180+ SCI journal and top-tier conference articles such as ICCV, ECCV, ICDE, IJCAI, TNNLS, TCYB, TIE, TII, TFS, etc. His  Google H-index is 40+ with about 7000 citations. He also serves as associate editors of  international SCI journals including IEEE Transactions on Emerging Topics in  Computational Intelligence, and Neurocomputing.


报告摘要:In autonomous localization, visual SLAM (Simultaneous Localization and Mapping), and VPR (Visual Place Recognition) are two modules providing prior location information for downstream tasks such as loop closure and global localization. For these reasons, they are two critical components for many applications such as robot navigation, autonomous driving, and augmented and virtual reality. Traditionally, the performance of these two modules relies on the accurate real-world input from expensive sensors (such as mechanical Lidar), causing high cost in massive production. To effectively reduce the cost, a fusion of multiple low-cost sensors is proposed in recent years, but the performance is relatively unsatisfactory. In this talk, we will introduce three of our latest works to tackle the performance issue of Visual SLAM and VPR in the manner of multi sensor fusion. Our experimental results show that the proposed methods can achieve SOTA performance with real-time inference under information fusion of multiple low cost sensors such as RGB camera, solid-state Lidar, and IMU.