A SURVEY ON DATA-DRIVEN TREE MODELING

PhD Qualifying Examination


Title: "A SURVEY ON DATA-DRIVEN TREE MODELING"

by

Miss Jinglu WANG


Abstract:

Trees are ubiquitous through the world and the virtual world which lacks 
vegetation appears lifeless and artificial. To achieve realism, significant 
progress has been made in modeling trees over the years. This paper surveys the 
state-of-the-art approaches for modeling trees, primarily classified as 
rule-based, sketch-based, scan-based and image-based approaches. The classes of 
techniques are not mutually exclusive and they can also be combined to use. 
Rule-based approaches make use of sets of generative rules or grammars while 
the other three approaches are related to the input data, that is 2D sketching, 
3D scan points and single image or multiple images. We favor the data-driven 
methods because such approaches have the better potential for producing 
realistic tree models. A data-driven tree modeling pipeline is introduced in 
detail with the comparison and analysis of the state-of-the-art techniques. We 
mainly focus on the procedure of the branch grow varying from particle tracing, 
non-parametric synthesis to point guided techniques. We also discuss the data 
preparation for image-based methods, semantic tree segmentation, since all 
these approaches contain the user invention. Finally, we present our method 
incorporating the fully automatic tree segmentation scheme which can facilitate 
the tree modeling process. Our method is evaluated on a variety of examples.


Date:			Monday, 19 August 2013

Time:                   3:00pm - 5:00pm

Venue:                  Room 3501
                         lifts 25/26

Committee Members:	Prof. Long Quan (Supervisor)
 			Prof. Chiew-Lan Tai (Chairperson)
 			Dr. Huamin Qu
 			Dr. Pedro Sander


**** ALL are Welcome ****