| Project/Area Number |
63880007
|
|---|
| Research Category |
Grant-in-Aid for Developmental Scientific Research (B).
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Informatics
|
|---|
| Research Institution | University of Tokyo |
|---|
Principal Investigator |
TAKAGI Mikio Inst. of Industrial Science The Univ. of Tokyo Professor, 生産技術研究所, 教授 (30013113)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KITSUREGAWA Masaru Inst. of Industrial Science The Univ. of Tokyo Associate Prof., 生産技術研究所, 助教授 (40161509)
|
|---|
| Project Period (FY) |
1988 – 1990
|
| Project Status |
Completed (Fiscal Year 1990)
|
| Budget Amount *help |
¥54,500,000 (Direct Cost: ¥54,500,000)
Fiscal Year 1990: ¥8,000,000 (Direct Cost: ¥8,000,000)
Fiscal Year 1989: ¥18,500,000 (Direct Cost: ¥18,500,000)
Fiscal Year 1988: ¥28,000,000 (Direct Cost: ¥28,000,000)
|
|---|
| Keywords | Remote Sensing / Parallel Processing / Database / Image Processing / 並列コンピュ-タと超高速デ-タベ-ス処理 / 気象衛星NOAA / 画像デ-タベ-ス / Meteorological Satellite NOAA / Image Processing / Image Database / 並列処理 Parallel Processing / 画像データベース / Prallel Processing |
|---|
| Research Abstract |
The objective of this research is to develop an integrated NOAA satellite imagery processing system which automatically receives the image data from the satellite and generates cariburated temperature map.
In order to speed-up the image delivery service to the users, we introduced a parallel processing approach for sattelite image processing. First we parallelized the radiometric correction routine, which can be processed almost independently pixel by pixel.
Linear speed up was obtained. Second the geometric correction routine was parallelized. This routine was not so easy as the radiometric one. We developed a novel elliptic model to caribulate the mapping error, through which we could obtain very high quality image. The mapping is performed by matching some characteristic points such as peninsula and gulf, which can be processed in parallel. After that the original map is decomposed into mesh of certain size and transformed to generate specific coordinated map. The size of mesh is a key parameter, which we finally obtained optimally through several experiments over a number of areas. The quantitative analysis of the correction error is also examined in detail. All the routine was implemented on the shared memory multi-processor system, sequent s81. We could attain higher performance on sequent than on mainframe computer which we have so far used.
|
