2022 Fiscal Year Research-status Report
Development of a high-resolution shoreline extraction technique by using convolutional neural network based on X-band SAR imagery
| Project/Area Number |
22K14327
|
|---|
| Research Institution | Tokyo University of Marine Science and Technology |
|---|
Principal Investigator |
呉 連慧 東京海洋大学, 学術研究院, 助教 (50907615)
|
|---|
| Project Period (FY) |
2022-04-01 – 2025-03-31
|
| Keywords | shoreline extraction / X-band SAR / UAV / neural network / coast protection |
|---|
| Outline of Annual Research Achievements |
The following achievements have been made this year. (1) A high resolution and efficient UAV-based shoreline extraction technique was developed. A series of filed observation was conducted to obtain shoreline positions and coast images. Then several image processing techniques were employed to develop the UAV-based shoreline extraction technique. Accuracy of the developed technique was confirmed by comparing it with the HandyGPS measured shoreline. (2)A shoreline extraction method was developed by using DeepLab v3+ (a convolutional neural network architecture) based on publicly available C-band SAR data (Sentinel-1, 10m pixel resolution). The accuracy of the method is confirmed to be less than 1 pixel when the wave condition at the SAR observation time is relatively calm.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
The developed UAV-based shoreline extraction technique can be used to obtain shoreline positions at the time of SAR observation, which is necessary for generating the training and testing dataset for neural network. In addition, the architecture of the convolutional neural network was determined and well examined for shoreline extraction based on X-band SAR imagery.
|
| Strategy for Future Research Activity |
Filed observations are expected to be carried out at the time of X-band SAR observation to obtain the necessary dataset for the neural network model. Then construction of the shoreline extraction technique by using convolutional neural network based on X-band SAR imagery will be conducted.
|
| Causes of Carryover |
The main reason for incurring amount is that SAR data was largely discounted.The incurring amount will be used for improving the computing environment for developing the neural network model.
|
