2005 Fiscal Year Final Research Report Summary
Active acquisition of character information by a controllable video camera
| Project/Area Number |
16500096
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | Tohoku University |
|---|
Principal Investigator |
OMACHI Shinichiro Tohoku University, Graduate School of Engineering, Associate Professor, 大学院・工学研究科, 助教授 (30250856)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
ASO Hirotomo Tohoku University, Graduate School of Engineering, Professor, 大学院・工学研究科, 教授 (10005522)
IWAMURA Masakazu Osaka Prefecture University, Graduate School of Engineering, Research Associate, 大学院・工学研究科, 助手 (80361129)
|
|---|
| Project Period (FY) |
2004 – 2005
|
| Keywords | Pattern recognition / Computer vision / Character recognition / Video camera / Mosaicing / Super resolution / Environment recognition / Robot vision |
|---|
| Research Abstract |
The aim of this research is to develop a method of actively acquiring character information in scene images with a video camera that is controllable by a computer. Detection methods of the area where characters exist and accurate recognition method of the character images were investigated. An object recognition method in the image was also investigated. The main results of this research are as follows.
1.Character recognition in scene images
It is difficult to extract the area where the character string exists in a scene image because of low resolution, changing in the lighting condition, etc. An algorithm for precise character extraction using color and luminance information is proposed. Moreover, a method for detecting and recognizing an isolated character is also investigated.
2.Object recognition in scene images
A method for recognizing objects in a scene image at high speed was investigated. The color histogram has been reported to be robust to the change and occlusion of the object compared with the other features. The proposed method achieves quick object detection by constructing color histograms efficiently. A technique for identifying a silhouette by representing the silhouette by a hierarchical graph was also investigated. Moreover, a novel approach of a template matching by representing an object image by a function was also investigated.
|
