1994 Fiscal Year Final Research Report Summary
Computer Vision Based on Computational Subsumption Model
| Project/Area Number |
05452355
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Tokyo Institute of Technology |
|---|
Principal Investigator |
SATO Makoto Tokyo Inst.of Tech., P.& I.Lab., Associate Prof., 精密工学研究所, 助教授 (50114872)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
TAKAMATSU Ryo Tokyo Inst.of Tech., P.& I.Lab., Research Assoc., 精密工学研究所, 助手 (20216782)
KAWARADA Hiroshi Tokyo Inst.of Tech., P.& I.Lab., Prof., 精密工学研究所, 教授 (00016776)
|
|---|
| Project Period (FY) |
1993 – 1994
|
| Keywords | Subsumption / Eary Vision / Viewpoint and Visual Field / Computer Vision |
|---|
| Research Abstract |
It is possible to consider that behavior of animal is layred from fundamental one to higher one during the evolutionary process. One of the approach to understand whole visual system is starting with making the fundamental model in common with many animal, and piling up higher level model gradually.
Detecting and tracking the moving object are some of the fundamental function of visual system. This function is essentinal for creatures to prey and reproduce.
It is significant how to decide and how to control optimal viewpoint and visual field during detecting and tracking target object. Many animal have the mechanism to decide optimal viewpoint and visual field based on a priori information about target object.
In this paper, we made the mathematical model of detecting and tracking target object, and investigated the optimality of deciding and controlling viewpoint and visual field.
The outline of the model is as follows ; we treat recognition of object as the measurement process of visual pattern to estimate parameters that describe the characteristics of object. The object image with noise is observed by measurement system that has viewpoint and visual field. The position and size of target object is estimated by minimum variance estimation. Deciding viewpoint and visual field is formulated as making measurement system that minimize estimation error based on a priori information about target object.
Model has these characteristics; characteristics in common with many animals taken into model has its origin in the fact that visual system of many animals are very similar though theembryological origin are far different. Describing measurement system, deciding and controlling the optimal viewpoint and visual field keeps consistency when extending model to more complicated function of visual system. Less quantity of operations to estimate position and size of target object makes it easy to actualize real time vision system.
|
