| Project/Area Number |
03452185
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
計測・制御工学
|
|---|
| Research Institution | University of Tokyo |
|---|
Principal Investigator |
FUJIMURA Sadao Univ. of Tokyo, Dept. of Math. Eng. & Inform. Phys., Prof., 工学部, 教授 (30010961)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HANAIZUMI Hiroshi Hosei Univ.,Industrial Eng., Assoc. Prof., 工学部, 助教授 (60143385)
KIYASU Senya Univ. of Tokyo, Dept. of Math. Eng. & Inform. Phys., Res. Assoc., 工学部, 助手 (20234388)
ITO Tadashi Univ. of Tokyo, Dept. of Math. Eng. & Inform. Phys., Res. Assoc., 工学部, 助手 (20223159)
ISHIKAWA Masatoshi Univ. of Tokyo, Dept. of Math. Eng. & Inform, Phys., Assoc. Prof., 工学部, 助教授 (40212857)
|
|---|
| Project Period (FY) |
1991 – 1992
|
| Project Status |
Completed (Fiscal Year 1992)
|
| Budget Amount *help |
¥1,300,000 (Direct Cost: ¥1,300,000)
Fiscal Year 1992: ¥1,300,000 (Direct Cost: ¥1,300,000)
|
|---|
| Keywords | Image Processing / Hierarchical / High-Dimensionality / Remote Sensing / Accuracy / Efficiency / 多次元画像 / 識別率 / 特徴空間 / 変化域抽出 / レジストレ-ション / アフィン変換 / 分類 |
|---|
| Research Abstract |
Objective: In the field of sensing technology, high resolution by a sensor and high dimensionality of data have been seeked. This brings us a vast volume (flood) of data: this is the case, especially for image data having great volume by themselves. In order to use the data effectively, we have to integrate the information from the multi-dimensional data. Our objective is to develop a method for integration of fusion of spatial information included in multi-dimensional image using hierarchical processing of data, and to unify the method to include other dimension (such as time or spectrum) than space. Resuts: (1) We developed a hierarchical method-for change detection by using multi-temporal images and confirmed its validity. We used two ways of approach. One detects changed areas by direct operation on multi- dimensional data. The other is so to speak "post-processing" of classified images and reduces classification errors by hierarchical processing. The classified images are the results of classification for the data obtained by multiple observation. (2) We discussed a method to extract altered areas taking the significance or the trend of change into consideration. Characterization of change using the feature space (the data without changed are plotted on the straight line in the feature space) enables us to extract it with ease. (3) We applied a hierarchical algorithm to sper-high-dimensional data. We have been developing an imaging spectrometer which is capable of yielding 512 channel data at a time. We developed an algorithm to extract significant features for discriminating objects by spectral structure. (4) we developed an automatic registration algorithm which saves time and laborious work. It uses affine transform for registration using spatial hierarchy (triangle area is devided int smaller triangles if the distortion is not removed by affine transform).
|
