1998 Fiscal Year Final Research Report Summary
A Study on Object Tracking by Hierarchical Computational Model
| Project/Area Number |
09650403
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
情報通信工学
|
|---|
| Research Institution | TOKYO INSTITUTE OF TECHNOLOGY |
|---|
Principal Investigator |
SATO Makoto Tokyo Inst.of Tech, P&I Laboratory, Prof, 精密工学研究所, 教授 (50114872)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
TAKAMATSU Ryo Saitama University, Faculty of Economics, Assoc.Prof., 経済学部, 助教授 (20216782)
KAWARADA Hiroshi Tokyo Inst.of Tech, P&I Laboratory, Prof, 総合理工学研究科, 教授 (00016776)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Keywords | computer vision / object tracking / robot vision / active vision / subsumption / task oriented vision / realtime system / moment feature |
|---|
| Research Abstract |
There are two indispensable features to realize vision-based human interface ; robustness under general environment and real time ability under restricted hardware resource. Conventionally, those features are thought to be trade-off. However, visual system of animals can achieve simple tasks sufficient for constructing human interface with restricted resource and considerable robustness. We hypothesize that the reasons why visual system of animal is successful in alleviating those trade-offs are to use global feature of restricted area and layered structure of sensory system.
The first year of the research plan, we constructed visual models of three layers of tasks ; discovering, tracking and recognizing thetarget.
As the global feature of restricted area, local moment, moment feature of image weighted by Gaussian function, is introduced. Since the local moment has the capability to suppress unnecessary information outside of the area, it can be applied effectively to an image containing much noises and obstacle objects. Comparing with the ordinary moment, computational cost of the local moment is fairly small and can be easily applied to real-time system.
This year, we developed two applications of those models. The first one is a pointing device based on tracking and recognizing user's hand. Without putting any marker-like reference points on a hand, a user can point at any position on the screen by moving the hand, and can manipulate objects on the screen in several ways. Another application is head posture estimation method using an oval marker attached on user's head. As the feature, moments of image up tosecond degree are used.
|
