| Item type |
紀要論文 / Departmental Bulletin Paper_08(1) |
| 公開日 |
2025-04-14 |
| タイトル |
|
|
タイトル |
AIと高解像度ドローン空撮画像を用いた樹木個体の識別 : 様々な樹種が混交した森林の構造解析を目指して |
|
言語 |
ja |
| タイトル |
|
|
タイトル |
Identify individual trees using AI in high-resolution aerial images taken by a drone : Aiming for structural analysis of forests with a mixture of various tree species |
|
言語 |
en |
| 言語 |
|
|
言語 |
jpn |
| キーワード |
|
|
言語 |
ja |
|
主題 |
人工知能 |
| キーワード |
|
|
言語 |
ja |
|
主題 |
ドローン |
| キーワード |
|
|
言語 |
ja |
|
主題 |
樹木 |
| 資源タイプ |
|
|
資源タイプ |
departmental bulletin paper |
| ID登録 |
|
|
ID登録 |
10.15045/0002000413 |
|
ID登録タイプ |
JaLC |
| 著者 |
友常, 満利
Contreras, Luis
小酒井, 正和
関川, 清広
武藤, ゆみ子
岡田, 浩之
|
| 抄録 |
|
|
内容記述タイプ |
Abstract |
|
内容記述 |
In this study, we used a drone to take aerial photographs of the Tamagawa Gakuen campus, which has various types of forests, and analyzed the images using AI to identify individual trees. The number of trees identified by each method (Urban Tree Detection and DeepForest) differed greatly. In Urban Tree Detection, the gap between the crowns of adjacent trees was determined to be the top of trees in many cases, and tree individuals were not properly identified. In the DeepForest, the number of trees identified differed depending on the patch (tree crown) size setting, and smaller patch size settings identified relatively more appropriate tree individuals than other methods. Following these results, it was determined that individual tree identification using AI is difficult at present, and among the tested methods, DeepForest is the method with the most potential, including future expandability. For practical individual identification in the future, it was considered important to 1) divide the area to be analyzed into areas of similar types as much as possible, 2) improve learning accuracy by acquiring training data in the forest to be analyzed, and 3) create a program that can analyze multiple images and additional information. |
|
言語 |
en |
| bibliographic_information |
ja : 玉川大学学術研究所紀要
号 30,
p. 95-100,
発行日 2025-03-15
|
| ISSN |
|
|
収録物識別子タイプ |
PISSN |
|
収録物識別子 |
1341-8645 |
| 出版タイプ |
|
| item_11_publisher_13 |
|
|
出版者 |
玉川大学 |
|
言語 |
ja |
11_2024_95-100.pdf (2.0 MB)
