Exteme Underwater Environment Sensing Using Acoustic Camera
Project/Area Number |
21J12362
|
Research Category |
Grant-in-Aid for JSPS Fellows
|
Allocation Type | Single-year Grants |
Section | 国内 |
Review Section |
Basic Section 20020:Robotics and intelligent system-related
|
Research Institution | The University of Tokyo |
Principal Investigator |
WANG YUSHENG 東京大学, 工学系研究科, 特別研究員(DC2)
|
Project Period (FY) |
2021-04-28 – 2023-03-31
|
Project Status |
Completed (Fiscal Year 2022)
|
Budget Amount *help |
¥1,500,000 (Direct Cost: ¥1,500,000)
Fiscal Year 2022: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 2021: ¥800,000 (Direct Cost: ¥800,000)
|
Keywords | 2d forward looking sonar / 3d reconstuction / deep learning / multi-path reflection / rendering / fiduicial marker / pose estimation |
Outline of Research at the Start |
極限水中環境において,一般な光学センサーなどによるセンシングすることが困難である.次世代音響カメラは濁った水においても鮮明な二次元画像を取得することが可能である.しかし,特殊な投影原理などにより,様々なタスクに直接運用することができない.特に,音響画像を用いて環境情報の収集と水中ロボットのナビゲーションする技術は重要視されている.本研究は水中ロボットによる構造物の点検のためのマーカシステムを提案する.デザインされたマーカにより,マーカと音響カメラの相対位置・姿勢を推定することが可能である.本システムにより,ロボットによる水中構造物の自動点検は可能となる.
|
Outline of Annual Research Achievements |
This year, several aspects of acoustic camera sensing in extreme underwater environments were achieved. The first is 3D reconstruction using two to three acoustic images. 3D reconstruction with a small number of acoustic images is a challenging task. An elevation plane sweeping stereo network was proposed. After training the network with a well-built dataset, it is possible to generate high-quality 3D model. This can be applied to underwater monitoring. The second is ground echo modeling, multi-path reflection may influence recognition results using acoustic images. By assuming the ground is flat, it is possible to explicitly model the ground echo. This is an important fundamental research in this field. The third is testing differentiable rendering in acoustic camera. This is the key to building the relationship between image simulation using computer graphics techniques and computer vision tasks like pose estimation and 3D reconstruction. It can be used to solve inverse rendering problems and be integrated into deep learning frameworks. Early results proved the feasibility of the method. A field experiment in water tank was also carried out for dataset collection. Three international conference papers were accepted on the three aspects.
|
Research Progress Status |
令和4年度が最終年度であるため、記入しない。
|
Strategy for Future Research Activity |
令和4年度が最終年度であるため、記入しない。
|
Report
(2 results)
Research Products
(9 results)