林地覆蓋變異(Forest Land Cover Changes, FLCC)產生之原因,包括崩塌、風災所造成之變異及林地違規開發使用之變異,FLCC不但影響森林的生態服務功能,亦影響林區管理之行政工作,因此有效掌握轄區之FLCC,為森林經營管理之重要工作。本研究利用多期衛星影像之波譜反射特性,探討遙測影像之前期處理對FLCC分析之影響;利用不同影像變異演算法,探討其對FLCC分析準確度之差異性;建立多期衛星影像之FLCC之作業流程,以利於有效的林地非法使用監測。在影像前期處理方面,本研究探討幾何糾正及輻射糾正對FLCC分析之準確度影響,結果顯示影像前期處理對於FLCC萃取之準確度雖然提升有限,但為FLCC分析之必要過程;利用常態化差值植生指標差值法(Difference Normalized Difference Vegetation Index, dNDVI)、主成份分析法(Principal Component Analysis, PCA) 、光譜變化向量分析(Spectral Change Vector Analysis, SCVA)及寫入儲存插入法(Write Function Memory Insertion,WFM),四種影像變異分析演算法以及兩種影像空間解析力(2 m融合影像、8 m多光譜影像),進行FLCC分析,並比較變異演算方法及不同影像空間解析力,對FLCC分析之準確度影響,結果顯示dNDVI 演算法準確度最高,且具有較佳的穩定性,而在空間解析力方面,由於椒鹽效應會造成8 m多光譜影像對於FLCC偵測準確度較2 m融合影像為佳。為區隔變異點所造成的原因,本研究以非監督分類法配合林地違規開發之行為模式,進行變異點的分類,結果顯示變異點依dNDVI值的高低分為5群其準確性最高。林地違規使用所產生之變異點,可依據農墾地之開發行為模式,設立篩選門檻值,利用研究區之相關的地理空間資訊,去除非人為因素所造成之FLCC,其結果有助於林政業務之應用。
Forest Land Cover Changes (FLCC) caused by landslide, typhoon or human-induced that affected the ecosystem services and forest management policy. How to efficiently get FLCC information is an important task for forest management. In this study, we use the spectral characteristics of multi-temporal satellite images to investigate the accuracy of FLCC analysis with image pro processing and different change detection algorithm; using multi-temporal satellite images to build the FLCC processes and to monitor the illegal forest landuse. In the pro processing of image, this study investigated the effects of geometric and radiation correction for FLCC analysis, results showed that the pro processing is limit to increase the accuracy of FLCC analysis, meanwhile that is a necessary process for FLCC analysis. Four change detection algorithms (dNDVI, PCA, SCVA, and WFM) will be used to compare the accuracy of FLCC analysis; results showed that the dNDVI has higher accuracy and stable than other algorithm. We integrate the illegal model of forest landuse and use unsupervised classification to classify the cause of FLCC; results showed that has a higher accuracy by the value of dNDVI to divide into five categories. FLCC caused by the illegal forest landuse, can be based on monitoring purposes to establish the threshold of illegal using model by geographic information system, and removal the non-human factors of FLCC, results will help to forest administration.