2021 Fiscal Year Final Research Report
Construction of Classification System of Wood Species by Cognitive Spectroscopy
| Project/Area Number |
19H03015
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 40020:Wood science-related
|
|---|
| Research Institution | Nagoya University |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
稲垣 哲也 名古屋大学, 生命農学研究科, 准教授 (70612878)
|
|---|
| Project Period (FY) |
2019-04-01 – 2022-03-31
|
| Keywords | 近赤外分光法 / ハイパースペクトラル画像 / CNN / 認識科学分析手法 |
|---|
| Outline of Final Research Achievements |
In this research, "Deep learning, which is machine learning using a multi-layered neural network, was applied to visible/near-infrared seamless hyperspectral data of wood, and a protocol that can automatically discriminate wood species is constructed and cognitive. The goal is to establish a new cognitive science analysis method that can be called a spectroscopic copy.
When tree species were discriminated using CNN based on visible images and near-infrared hyperspectral images of 38 hardwood species, the correct answer rate when using near-infrared images reached 90.5%. This made it possible to show the possibility of a new cognitive chemical analysis method of "analyzing molecular vibration information on the sample surface and its spatial distribution by CNN".
|
| Free Research Field |
木質科学
|
| Academic Significance and Societal Importance of the Research Achievements |
本手法によって近赤外領域・可視領域両方での木材樹種判別の可能性を示した。顕微鏡写真からのバッチ抽出およびトレインデータ・テストデータの選択方法によって樹種判別の推定精度が大きく異なることが明らかとなった。以上一連の研究によって本手法の木材樹種判別への有効性および限界を確認した。また提案手法を様々な農産物評価にも応用し、その有用性を確認した。
|
