Visual Localization and Mapping in Dynamic Environments: A Survey

PhD Qualifying Examination


Title: "Visual Localization and Mapping in Dynamic Environments: A Survey"

by

Mr. Huajian HUANG


Abstract:

Visual localization and mapping are essential techniques that seek to enable 
the agent to understand surroundings via cameras and infer their poses with 
respect to reconstructed models. With the development of computer vision and 
the increase of computing power, they have made an impact in various 
applications, such as augmented reality, robot navigation, autonomous vehicles 
and etc. Under the assumption that the working environment is static, some 
impressive and promising results have achieved in these areas. However, when 
considering practical deployment in the real world, it is inevitable to 
encounter dynamic objects and illumination. Due to false correspondences or 
occlusion of previously tracked features, conventional frameworks would suffer 
performance degradation or even break down in dynamic environments. The 
capability of efficiently handling dynamic environments is key to the next 
generation of visual localization and mapping.

In this survey, we review the development of visual localization and mapping 
and investigate existing approaches taking into account dynamic environments. 
The discussion concentrates on three concepts: how to perform robust 
localization and mapping, how to achieve object-level tracking and 
reconstruction, and how to effectively maintain 4D representations of 
environment suitable for spatial intelligence. The article ends up with the 
proposal of future directions.


Date:			Friday, 30 July 2021

Time:                  	4:00pm - 6:00pm

Zoom meeting:
https://hkust.zoom.com.cn/j/98844418126?pwd=YTVnSWM5R3o5dVJiekZzVmpKQ2orQT09

Zoom meeting venue: 	Room 5564
 			Lifts 27/28

Committee Members:	Dr. Sai-Kit Yeung (Supervisor)
 			Dr. Qifeng Chen (Supervisor)
 			Prof. Pedro Sander (Chairperson)
 			Prof. Chi-Keung Tang


**** ALL are Welcome ****