1999 Fiscal Year Final Research Report Summary
A realistic model generation for virtual world construction using real-time image sequence
| Project/Area Number |
09480069
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | KYUSHU UNIVERSITY |
|---|
Principal Investigator |
TANIGUCNI Rin-ichiro Graduate School of Information Science and Electrical Engineering, Kyushu University, Professor, 大学院・システム情報科学研究科, 教授 (20136550)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KUSAKA Shigeru Graduate School of Information Science and Electrical Engineering, Kyushu University, Associate Professor, 大学院・システム情報科学研究科, 助教授 (70234416)
SUGANUMA Akira Graduate School of Information Science and Electrical Engineering, Kyushu University, Associate Professor, 大学院・システム情報科学研究科, 助教授 (70235852)
KAMATA Seiichiro Graduate School of Information Science and Electrical Engineering, Kyushu University, Associate Professor, 大学院・システム情報科学研究科, 助教授 (00204602)
TSURUTA Naoyuki Faculty of Engineering, Fukuoka University, Lecturer, 工学部, 講師 (60227478)
ARITA Daisuke Graduate School of Information Science and Electrical Engineering, Kyushu University, Research Associate, 大学院・システム情報科学研究科, 助手 (70304756)
|
|---|
| Project Period (FY) |
1997 – 1999
|
| Keywords | Virtual World / Model Generation / Computer Vision / Image Sequence Analysis / Multi-view Analysis / Shape Analysis / Motion Analysis |
|---|
| Research Abstract |
(1) Prototype system development of real-world and virtual-world fusion
In the framework that required information to present the objects in the virtual space consists of a priori information and a posterior observation, we have established the basic methodology of 3D model generation, of which features are as follows : (1) a posteriori observation is not very difficult, (2) a priori information is easily supplied to the system, (3) complex objects, or multi-part deformable objects such as humans and animals. The method is called Analysis by image synthesis, in which model parameters are modified so that the error between the synthesized image based on the model parameters and the observed image is minimum. As the geometrical model to represent object parts, we have used Deformable Super Quadrics (DSQ), which can represent tapering and bending of the parts. Basic modules included the system as follows :
・Tracking system of multi-part deformable objects based on Analysis by image synthesis method
・GUI-based model generation system for initial 3-d objects, which is used in the tracking system
・Visualization of acquired object models and texture mapping
(2) Speed-up and accuracy improvement of the system
To improve the system development in 1. We have investigated the following issues.
・To achieve the speed-up of the system, we have developed time model-space gradient method, which linearize the system mode, and which can be solved by solving a linear simultaneous equation.
・An algorithm of view-point selection among multiple views to achieve more accurate analysis has been developed.
|
