1996 Fiscal Year Final Research Report Summary
Fundamental Analytic of Radar Imagery
| Project/Area Number |
06650589
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
交通工学・国土計画
|
|---|
| Research Institution | KYOTO UNIVERSITY |
|---|
Principal Investigator |
OKAMOTO Atsushi Kyoto University, Graduate School of Engineering, Associate Professor, 工学研究科, 助教授 (30026265)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
ONO Tetsu Kyoto University, Graduate School of Engineering, Instructor, 工学研究科, 助手 (00263103)
HATTORI Susumu Fukuyama University, Faculty of Engineering, Associate Professor, 工学部, 助教授 (50108122)
|
|---|
| Project Period (FY) |
1994 – 1996
|
| Keywords | radar imagery / stereo matching / simulated annealing / re-extraction of training data / analysis of different kinds of images |
|---|
| Research Abstract |
(1) For real-time production of topographic maps from stereo or triplets radar images, stereo matching is required to be automated. However, the stereo matching methods used at present have rather low accuracy. Thus, we investigated this problem fundamentally and developed a new method of stereo matching based on simulated annealing. Tests with many simulated examples clarified that the proposed method has fairly high accuracy.
(2) Acquisition of training data is a very tedious work in superwised classification using iamge data. A land use map may be very effectively employed for this purpose by removing the outliers. We developed a blunder detection method based on simulated annealing. Through this trainning data cleaning the classification accuracy was improved by 5 to 10 percents.
(3) The simultaneous analysis of air-borne radar imagery and aerial photographs strengthens the geometry of radar imagery. However, the geometrical properties of the orientation problem of radar-and central-perspective images have not been fully clarified. The theoretical and practical investigation revealed that air-borne radar images can be anlysed very accurately by treating them together with other images having strong geometry.
|
