Please use this identifier to cite or link to this item:
標題: 輻射同態化對SPOT衛星影像用於變遷偵測影響之研究─以合歡山地區為例
The Effecct of Radiometric Normalization on SPOT Satellite Image Change Detection
作者: 何珮艷
Her, Pey-Yann
關鍵字: 衛星影像相對輻射糾正
Relative Radiometric Normalization of Satellite Images
Pseudo Invariant Feature
Linear Regression Normalization
Histogram Matching
Normalized Difference Vegetation Index
Root Mean Square Error
Change Detection
出版社: 森林學系
摘要: 台灣之山區佔全島面積之比例超過50%,且因高山植群對環境變化甚為敏感,故此地區地被變遷之偵測日益重要。以衛星影像進行地被植群之變遷偵測可節省實地調查經費與人力,且利於時序數值影像之處理與儲存。然因大氣、日照、感測儀及物候等因素導入不同時期影像上的輻射差異,降低變遷偵測之正確率。本研究探討上述輻射差異對變遷偵測之影響,以確認影像同態化對變遷偵測之重要性,且評估影像同態化方法之成效。直方圖匹配、線性迴歸及規整差植生指標應用在合歡山地區分屬1994與1998年四季之八期SPOT影像上,並評比其效果。就影像外觀、均方根誤差、變遷偵測試驗及操作難易來看,直方圖匹配之同態化效果最優,規整差植生指標次之,線性迴歸殿後,且後兩者效果不穩定。就波段來看,紅光同態化效果優於綠光,兩者皆優於近紅外光。同態化成效傾向隨影像對間季差縮小而提高,亦即影像對屬同季,特別是夏季,優於分屬冬夏者。變遷偵測率高低之關鍵在無變遷誤授至疑似或有變遷,故影像同態化對變遷偵測甚為重要。
The rate of mountain area is over 50% in Taiwan. Landuse change monitoring has become important increasingly, because alpine vegetation is subject to environmental change. Change detection through satellite images may save much cost and labor, and digital images also facilitate processing and storage in a time?series database. However, radiometric difference introduced by atmosphere, illumination, sensor, and plant phenology may reduce the accuracy of change detection. The study was intended to confirm the importance of image normalization. It also eevaluated the performance of different image normalization methods in relation to change detection. Histogram matching (HM), linear regression normalization (LRN), and normalized difference vegetation index (NDVI) were applied to eight?date SPOT images in four seasons of 1994 and 1998 in Hohuan mountain area. The methods were compared in terms of their capability to improve visual image quality, statistical robustness, and ease of implementation. HM exhibited the best overall performance among them, NDVI was second to HM, and LRN was the last. Image normalization had better performance using visible bands than using near?infrared band. The performance had a tendency to increase with a smaller season difference between an image pair. Namely, an image pair acquired in the same season, particulary in summer, was better than one acquired in winter, the other in summer or vice versa. The key to a high or low accuracy in change detection was the errorneous assignment of no?change pixels to the likely?change or true?change pixels. Thus image normalization was of great important for change detection.
Appears in Collections:森林學系



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.