查詢結果分析
來源資料
頁籤選單縮合
| 題 名 | 應用深度學習方法於大範圍山崩自動判釋=Application of Deep Learning Methods for Wide-area Automated Landslide Detection |
|---|---|
| 作 者 | 簡留玄; 余執彰; 黃春銘; 吳庭瑜; 陳德偉; 謝有忠; | 書刊名 | 中興工程 |
| 卷 期 | 168 2025.07[民114.07] |
| 頁 次 | 頁(F)1-(F)9 |
| 分類號 | 312.831 |
| 關鍵詞 | 深度學習; 卷積神經網路; 山崩判釋; Deep learning; CNN; Landslide detection; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 山崩調查對坡地整治規劃至關重要。近年來,隨著遙測技術的發展,人力難以到達的偏遠山區也能透過遙測 技術有效揭露山崩現況,進而有助於臺灣大範圍山崩調查工作。過去山崩調查主要仰賴人力判讀衛星影像,耗時 費力,隨著機器學習與深度學習的興起,這些技術已延伸至衛星影像、航照等遙測影像的分析,使得即時且廣域 的山崩調查成為防災工作的重要目標之一。遙測影像與一般影像有著不同的特性,因具有空間資訊,得以疊加不 同種類的資料進行分析,過去的研究除了利用影像外,也加入了地形資訊以排除誤判。然而,山崩具有自然物不 規則的特徵,並受地質、地形的影響,因此在不同區域有著差異,為了降低區域差異對模型的干擾,本研究將利 用深度學習模型,以影像與坡度作為模型訓練資料,並進行分區訓練。 |
| 英文摘要 | Landslide investigations are crucial for planning and implementing slope stabilization projects. Recent advances in remote sensing technologies have enabled effective detection and study of landslides in remote mountainous areas that were previously inaccessible. While traditional manual interpretation of satellite imagery was labor-intensive and time-consuming, the emergence of machine learning and deep learning has made automated analysis possible. These techniques can process remote sensing imagery—including satellite and aerial photographs—enabling real-time, wide-area landslide detection for disaster risk reduction. The key advantage of remote sensing imagery lies in its spatial information, enabling integration and analysis across multiple data layers. Previous studies have reduced false positives by combining visual imagery with topographic data. However, landslides are irregular natural phenomena shaped by local geological and topographic conditions, leading to regional variations that might affect model performance. This study addresses these spatial differences and improves model performance through a deep learningñ based approach, combining imagery and slope data with a zonal training strategy. |
本系統中英文摘要資訊取自各篇刊載內容。