應用多時期遙測影像於宜蘭海岸土地覆蓋變遷分析之研究

Application of Multitemporal Remote Sensing Images in Analyzing Land Cover Change along Yilan Coast

吳治達(Chih-Da Wu) ； 詹進發(Jihn-Fa Jan) ； 徐郁晴(Yu-Ching Hsu)

《工程環境會刊》 27期 (2011/07) Pp. 93-104

https://doi.org/10.6562/JEE.2011.27.6

摘要

衛星影像可迅速蒐集大範圍的地表資料，多時期衛星影像可提供土地覆蓋變遷分析所需之有力資訊，故本文以宜蘭海岸地區2003年、2006年與2009年之SPOT-5影像為研究材料，探討土地覆蓋變遷的趨勢。首先，以監督式影像分類法將影像分類為水體、建成與交通用地、沙地、農地、林地等五種土地覆蓋類別，以影像分類之結果為基礎，比較三個時期之土地覆蓋變遷；其次，參考前人對宜蘭海岸變遷之研究，選擇可能影響宜蘭海岸變遷的自然與人為因子，以logit迴歸模式進行分析，利用2003年與2006年影像分類成果建立解釋模式，以推估未來各種土地覆蓋類別之空間分佈的情形；最後以2009年影像分類成果進行驗證，結果顯示模式推估的整體準確率可達80%以上。

關鍵字

SPOT-5衛星影像 ； 監督式影像分類法 ； 土地覆蓋變遷 ； 二項式Logit迴歸

並列摘要

Satellite images are capable of rapidly collecting data about large-scale earth surface, and multitemporal remote sensing images can provide forceful information for analyzing land cover changes over time. This study used three SPOT-5 images acquired between 2003 and 2009 to investigate the trend of land cover change in the coastal zone of Yilan. First, supervised classification method was used to classify the images into 5 land cover types, including water body, built-up areas, sandy area, cropland, and forest land. Based on the classification results, the land cover changes among these three periods was compared. Secondly, by referring to previous studies on Yilan coastal changes, this study selected several natural and human factors that are likely to affect coastal changes to build a logistic regression model for representing land cover change between 2003 and 2006. The model was then used to evaluate spatial distributions of various land cover types. Finally, the classification results of 2009 image was used to validate predictions from the logit model. The results indicate that overall accuracy of the prediction was above 80%.

並列關鍵字

SPOT-5 satellite image ； supervised image classification method ； land cover change ； binomial logit model