A SURVEY ON OUTDOOR POINT CLOUD SEGMENTATION

PhD Qualifying Examination


Title: "A SURVEY ON OUTDOOR POINT CLOUD SEGMENTATION"

by

Mr. Maosheng YE


Abstract:

Despite the great progress in vision tasks in the 2D image domain, point 
cloud learning still faces many challenges. Due to its sparsity and 
irregularity, directly applying the methods used in the 2D tasks is not 
feasible and applicable, especially considering the outdoor scenarios with 
more than 100K points. Several approaches have been proposed based on 
various representations to explore better feature modeling ways. The key 
idea of these methods is to convert the point cloud into one or several 
representations and then propose different architectures or operations for 
context information learning since each representation has its own merits 
for data processing.

In this survey, we present a comprehensive review of large-scale outdoor 
scenario point cloud segmentation, which has been a crucial task for 
autonomous driving and 3D reconstructions. We first introduce the 
background and current state-of-the-art methods to provide an overall view 
of this task. Then we divide the methods into single-representation and 
multi-representation categories and discuss the limitations of the 
corresponding methods. Next, self-supervised and semi-supervised 
approaches are also introduced. Finally, we conclude the survey with some 
future research directions.


Date:  			Thursday, 19 May 2022

Time:                  	10:00am - 12:00noon

Zoom Meeting: 		https://hkust.zoom.us/j/7575421761

Committee Members:	Dr. Qifeng Chen (Supervisor)
 			Dr. Hao Chen (Chairperson)
 			Prof. Chiew-Lan Tai
 			Dr. Dan Xu


**** ALL are Welcome ****