2022 Fiscal Year Final Research Report
Developing Mathematical Methods for Applying Hyperspectral Imaging to Earth Observation
| Project/Area Number |
20K11951
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 61030:Intelligent informatics-related
|
|---|
| Research Institution | Shizuoka University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | ハイパースペクトルイメージング / ミクセル分解 / 非負行列分解 / スペクトル法 / 凸最適化 |
|---|
| Outline of Final Research Achievements |
This research project aimed to enhance the performance of algorithms for unmixing hyperspectral images. In particular, we studied Hottopixx method and spectral method, which are promising approaches to hyperspectral unmixing. Hottopixx method was shown to be robust to noise. But there are two issues: one is that the method requires us to estimate the noise involved in the data matrix before running; and another is that the computational cost is expensive. We revised the algorithm of Hottopixx method, and overcame the issues. In COLT 2015, Peng et al. showed the performance of spectral method. We improved their results by enhancing the structure theorem they showed. Based on the results we obtained, we developed a novel algorithm for spectral method, and studied the performance in theory and practice.
|
| Free Research Field |
知能情報学
|
| Academic Significance and Societal Importance of the Research Achievements |
ハイパースペクトルイメージングは地球表面の観測に活用されている．植生分布の把握や海洋汚染調査などを人が直接実施すると大きなコスト（労力や時間など）が伴うが，人工衛星に搭載されたハイパースペクトルセンサを用いると，一度の観測で広域な領域を調べることができる．そのためコストを大幅に軽減することが可能となる．ハイパースペクトルセンサで取得した画像から，端成分スペクトルと含有率を求めることをミクセル分解と呼ぶ．ミクセル分解はハイパースペクトルイメージングを活用するための基本的な問題である．本研究ではミクセル分解手法のアルゴリズムを改良し，その有効性を明らかにした．
|
