Deep forest for disaster monitoring with multi-temporal and multi-modal earth observation data
| Project/Area Number |
19K20309
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 61010:Perceptual information processing-related
|
|---|
| Research Institution | Institute of Physical and Chemical Research |
|---|
Principal Investigator |
XIA JUNSHI 国立研究開発法人理化学研究所, 革新知能統合研究センター, 研究員 (00830168)
|
|---|
| Project Period (FY) |
2019-04-01 – 2022-03-31
|
| Project Status |
Completed (Fiscal Year 2021)
|
| Budget Amount *help |
¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
Fiscal Year 2021: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2020: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2019: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
|
|---|
| Keywords | disaster monitoring / multi-modality / deep forest / deep learning / Damage mapping / Machine learning / Multi-source datasets / Deep forest / Siamese networks / Disaster monitoring / Multi-temporal / Multi-modal / Remote sensing |
|---|
| Outline of Research at the Start |
The outline of research is divided into three parts:
1) constructing the database of disasters;
2) developing Deep forest for disaster monitoring;
3) developing Deep transfer forest among different disasters.
|
| Outline of Final Research Achievements |
We developed methods for disaster damage mapping by using multi-temporal and multi-modal earth observation data and deep learning methods (e.g., CNN, deep forest, PU learning). We also studied different-modality learning: how to learn relevant features between two modalities. Our deep learning methods were applied to natural disaster mapping, such as the 2018 Sulawesi, the Sunda Strait earthquake/tsunami, etc., which can solve the limited training sample problem and outperform the traditional techniques.
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究において、様々なセンサーから得られた地球観測データを用いて災害マッピングを調査できること、また深層学習手法を用いて分類能力を高めることの必要性について証明しました。実際には、教示データの欠如，観測データの不均一性，学習方法の能力などの次の課題により，災害被害マッピングの予測結果が低くなります。最も顕著な成果は，提案された方法がパフォーマンスを改善することが可能になったということです。
|
Report
(4 results)
Research Products
(12 results)
