本計畫完成2009年~2015年之10幅福衛二號衛星影像之崩塌面積判釋

本計畫完成2009年~2015年之10幅福衛二號衛星影像之崩塌面積判釋、統計及比較。配合遙測影像識別所需樣本，完成7次現地調查，對Pleiades衛星影像進行自動分類的精確度探討及分析。配合另案計畫所獲得之2015年航拍影像，完成崩塌裸露地數量、範圍及大小統計。導入加值影像分類法，分類精確度提升至98%，顯示此一方法可以大幅提升分辨率。。

Results (English) 1:

2009 ~2015 this plan was completed years of catchment area of 10 Fuwei second satellite image interpretation, statistics and more. Samples needed to work with remote sensing image recognition, 7 times the field survey, the accuracy of the automatic categorization of the Pleiades satellite image analysis and discussion. Combined with the other case plan obtained by 2015 years of aerial images, complete collapse exposed the number, scope and size statistics. Import bonus image classification accuracy up to 98%, shows this approach can dramatically improve the resolution ...

Results (English) 2:

The project completed in 2009 to 2015 of 10 Interpretable collapse area of ​​FORMOSAT two satellite images, analyzed and compared. Image recognition with telemetry samples needed to complete the seven current investigation, for Pleiades satellite images for the accuracy of automatic classification of Discussion and Analysis. Lingan plan with the aerial images of the 2015 acquisition, completed collapses exposed to the number, scope and size statistics. Importing bonus image classification method, classification accuracy increased to 98 percent, indicating that a method can significantly improve the resolution. .

Results (English) 3:

The plan completed in 2009 to 2015 10 pieces FORMOSAT release, statistics and comparison of two satellite images of landslide area classification. With remote sensing image recognition of the required sample and 7 times in situ investigation is completed, the Pleiades satellite images for the accuracy of the automatic classification discussion and analysis. With Lingan meter painting 2015 aerial image, complete collapse bare quantity, scope and size statistics. Import value added image classification, classification accuracy increased to 98%, showing the method can significantly improve the resolution.