2022 Fiscal Year Final Research Report
Prediction of road damage based on low-cost, high-volume, and high-frequency data accumulation using deep learning
Project/Area Number |
20K14799
|
Research Category |
Grant-in-Aid for Early-Career Scientists
|
Allocation Type | Multi-year Fund |
Review Section |
Basic Section 22010:Civil engineering material, execution and construction management-related
|
Research Institution | The University of Tokyo |
Principal Investigator |
Maeda Hiroya 東京大学, 生産技術研究所, 特任研究員 (90853200)
|
Project Period (FY) |
2020-04-01 – 2023-03-31
|
Keywords | 土木 / 舗装 / 画像処理 / AI / インフラメンテナンス |
Outline of Final Research Achievements |
In this research, we conducted research to automatically detect road damage locations such as cracks and holes using only widely used hardware such as smartphones and drive recorders. Furthermore, we collected road damage data not only in Japan but also in India and the Czech Republic, and built an automatic detection model that can be applied in any country. At that time, by tuning the automatic detection model created with Japanese road data, it was shown that the automatic detection model in India and the Czech Republic can be created with a small amount of training data.
|
Free Research Field |
土木
|
Academic Significance and Societal Importance of the Research Achievements |
道路メンテナンスは人手不足、予算不足が深刻であり、従来のように人手や高価な専用車両を用いた点検を継続的、網羅的に実施していくことが難しくなっている。このような状況で、本研究ではスマートフォンやドライブレコーダーといった比較的安価な機材のみを用いて、低廉迅速に道路損傷データを収集できることを示し、社会的な意義が大きいと考えている。また、日本国内で作成した損傷の自動検出モデルを海外で適用することができる可能性を示し、複数の国における道路損傷データを整備、公開したことは学術的な意義が大きいと考える。
|