查詢結果分析
來源資料
頁籤選單縮合
| 題 名 | 基於UAV影像之都市土地利用與覆蓋的分類：監督式與非監督式機器學習之評估=Classification of Urban Land Use and Land Cover Based on UAV Imagery: Evaluation of Supervised and Unsupervised Machine Learning Methods |
|---|---|
| 作 者 | 范慶龍; 閻善民; | 書刊名 | 臺灣水利 |
| 卷 期 | 73:2=290 2025.06[民114.06] |
| 頁 次 | 頁1-23 |
| 分類號 | 312.831 |
| 關鍵詞 | 機器學習; 深度學習; UAV影像; 土地利用和覆蓋; Machine learning; Deep learning; UAV imagery; Land use and land cover; |
| 語 文 | 中文(Chinese) |
| DOI | 10.6937/TWC.202506_73(2).0001 |
| 中文摘要 | 都市土地利用與覆蓋分類是遙測技術中的一個關鍵任務，它提供了關於都市區域內不同土地使用類型分佈的重要資訊。本研究使用無人機獲取高解析度影像，並評估監督式與非監督式機器學習方法在進行都市土地利用與覆蓋的分類性能。透過混淆矩陣計算整體準確度、F1 Score等評估指標，結果顯示監督式機器學習的分類效果優於非監督式機器學習。特別是深度學習模型DeepLab V3和U-Net的整體準確度（Overall Accuracy, OA）分別達到92.3%和91.8%，在建築物、道路、樹木及草地的分類中表現突出；傳統機器學習模型（SVM和KNN）的OA分別為85.4%與84.8%，表現次之；非監督式方法（Clustering）的OA僅為79.4%，在樹木及草地的分類效果較差。 |
| 英文摘要 | Urban land use and land cover classification is a critical task in remote sensing, providing essential information about the distribution of various land use types within urban areas. This study utilizes high-resolution images acquired from Unmanned Aerial Vehicles (UAVs) to evaluate the performance of supervised and unsupervised machine learning methods in urban land use and land cover classification. Using confusion matrix-derived metrics such as Overall Accuracy (OA) and F1 Score, the results indicate that supervised machine learning methods outperform unsupervised methods. In particular, deep learning models such as DeepLab V3 and U-Net achieved OA values of 92.3% and 91.8%, respectively, excelling in the classification of buildings, roads, trees, and grasslands. Traditional machine learning models, such as SVM and KNN, demonstrated slightly lower performance with OA values of 85.4% and 84.8%, respectively. In contrast, unsupervised methods, such as clustering, achieved an OA of only 79.4%, struggling particularly in the classification of trees and grasslands. |
本系統中英文摘要資訊取自各篇刊載內容。