無人飛行載具產製可見光與近紅外光數值地形分析崩塌潛感值：以神木村地區為例

Determination of Landslide Susceptibilities Using UAV-Borne RGB and NIR images: A Case Study of Shenmu Area in Taiwan

蕭宇伸(Yu-Shen Hsiao) ； 莊大賢(Ta-Hsien Chung) ； 陳樹群(Su-Chin Chen) ； 張榮傑(Jung-Chieh Chang)

《中華水土保持學報》 49卷3期 (2018/09) Pp. 142-153

https://doi.org/10.29417/JCSWC.201809_49(3).0002

摘要

本文利用無人飛行載具建置地形，以研究台灣神木村地區崩塌潛感相關課題。經由搭載可見光與近紅外光相機的無人飛行載具航拍，獲得神木村周邊10公分空間解析度的正射鑲嵌影像與50公分網格解析度的數值地表模型。本研究利用GPS衛星控制點檢核數值地表模型，並利用羅吉斯迴歸分析神木村山崩潛感值，選用包括高度、坡度、坡向、地形起伏度、地形粗造度與總曲率等6個崩塌潛感因子。本研究主要相關結果如下：(1)可見光產製與近紅外光產製的數值地表模型，高程精度分別為0.953公尺與2.236公尺；(2)坡度與坡向為本研究區域之最重要崩塌潛感因子；(3)可見光影像產製之數值地形，比近紅外光更適合應用於山崩潛感值預估。

關鍵字

無人飛行載具 ； 崩塌潛感 ； 羅吉斯迴歸

並列摘要

This study conducted a landslide susceptibility analysis at Shenmu area in Taiwan by using terrain models constructed using unmanned aerial vehicles (UAVs). High-resolution orthomosaics and digital surface models (DSMs) were both obtained from several UAV practical surveys by using red-green-blue (RGB) and near-infrared (NIR) cameras, respectively. GPS control points were used for evaluating the DSMs. The algorithm for landslide susceptibility prediction is based on logistic regression, in which elevation, terrain slope, terrain aspect, terrain relief, terrain roughness, and curvature, were the primary factors. The results are as follows: (1) the vertical accuracies of RGB- and NIR-derived DSMs are 0.953 and 2.236 m, respectively; (2) terrain slopes and aspects are the most influential factors in landslide susceptibility prediction; (3) the DSM derived from RGB images are more appropriate for landslide susceptibility prediction than that from NIR.

並列關鍵字

UAV ； Landslide Susceptibility ； Logistic Regression