2019 Fiscal Year Annual Research Report
System for Automatic and Real-time Generalization of Catastrophe Maps based on Deep Learning Methods
| Project/Area Number |
19J13500
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
郭 直霊 東京大学, 新領域創成科学研究科, 特別研究員(DC2)
|
|---|
| Project Period (FY) |
2019-04-25 – 2021-03-31
|
| Keywords | Deep Learning / Super Resolution / Remote Sensing |
|---|
| Outline of Annual Research Achievements |
We address the challenging task of the semantic segmentation of land features via multi-source remote sensing imagery with different spatial resolutions. Unlike previous works that mainly focused on optimizing the segmentation model, we propose to integrate super-resolution techniques with the existing framework to enhance the segmentation performance. The results confirmed that the proposed method is a viable tool for building semantic segmentation, especially when the resolution is unaligned.
Publised papers:
1. Super-Resolution Integrated Building Semantic Segmentation for Multi-Source Remote Sensing Imagery
2. GeoSR: A Computer Vision Package for Deep Learning Based Single-Frame Remote Sensing Imagery Super-Resolution
|
| Current Status of Research Progress |
Current Status of Research Progress
3: Progress in research has been slightly delayed.
Reason
The data source we use is based on satellite and aerial imagery, UAV has not been applied yet.
|
| Strategy for Future Research Activity |
1. System testing
Carefully test the obtained system by considering the feasibility, stability, accuracy and robustness.
2. Practical application
Apply this system in real catastrophe condition if it’s possible, and contribute to real rescues.
|
