多光譜影像與空載光達於山崩植生復育監測之研究

Landslides Vegetation Restoration Monitoring by Using Airborne LiDAR and Multispectral Imagery

楊孟學(Mon-Shieh Yang) ； 吳銘志(Ming-Chee Wu) ； 劉進金(Jin-King Liu)

《航測及遙測學刊》 20卷3期 (2016/05) Pp. 171-181

https://doi.org/10.6574/JPRS.2016.20(3).2

摘要

邊坡穩定往往與地形、地質及植生息息相關，這些條件也常被用來作為評估邊坡穩定的因子。然而對於大範圍邊坡穩定之調查，往往因面積過大而使得調查較為困難。隨著遙測技術的發展，近年來常藉由遙測影像輔助山區災害的調查。本研究結合多光譜遙測影像以及空載光達產製之地表粗糙度影像建立一適於山崩植生復育調查之三維植生覆蓋指標，該指標同時具有二維的植生健康度訊息以及三維的生長維度訊息，期藉此指標能更有效的評估植生的健康度與復育生長情形。其研究結果也顯示，整合多光譜影像所提供的二維資訊與光達三維資訊之三維植生指標，該指標除了整體趨勢與常用的植生指標趨勢一致外，對於植生之垂直維度的生長特徵也更加詳細。以分類精度比較，NDVI 輔助SPOT 影像之分類成果精度為72.61%、CHM 輔助SPOT 影像之分類成果精度為77.04%、三維植生覆蓋指標輔助SPOT 影像之分類成果精度達89.75%。此外，利用三維植生指標輔助影像自動化判識之精度也有所提升，顯示三維植生指標可作為山崩植生復育重要的評估指標。

關鍵字

光達 ； 地質災害 ； 地表粗糙度

並列摘要

Slope-safety factors associated with topography, geology, and vegetation, which can be considered as indices of the parameters of safety factors, is often quantified to assess their contributions to landslides. However, for large areas, variations in the parameters included for analyzing the safety factors are too large to quantify accurately. With advances in the efficiency and accuracy of investigative techniques, remote sensing technologies have become widely employed in assessing landslide hazards. In this paper, the effectiveness of using airborne laser altimeter retrieved data to evaluate landslide vegetation restoration is analyzed. For this purpose, satellite multi-spectrum image were integrated with airborne laser altimeter surface-roughness data to develop a method to characterize vegetation restoration on landslide areas. The analysis was performed in both 2D and 3D scales, the results of the integrated method suggest that airborne laser altimeter data can be used effectively for evaluating landslide vegetation restoration than using the NDVI, besides, the 3D vegetation index also shows the same characteristics and can describe the features of the vegetation effectively than using spectrum information only; the amplitude of the 3D vegetation index represents the vegetation condition both in the vertical and the horizontal dimension. The classification accuracy is improved from 72.61% for using traditional multispectral classification to 89.75% when using 3D vegetation index integrated with SPOT images. The results of the integrated method suggest that integrated method can be used effectively for evaluating landslide vegetation restoration.

並列關鍵字

airborne LiDAR ； geological hazard ； surface roughness