多期卫星影像相对辐射纠正於台中乌石坑地区变迁侦测之探讨.PDF

林業研究季刊 28(1) ：29-38, 2006 29 研究報告 多期衛星影像相對輻射糾正於台中縣烏石坑地區 變遷偵測之探討 潘麗慧 1 羅南璋 2 黃凱易 3 【摘要】多期影像變遷偵測常應用於地景變遷監測，但因非地表因子如大氣效應、光照、觀視幾 何和感測儀校準及地表因子季節物候和地形效應等形成假變遷，以致變遷偵測正確率降低。本研 究探討非地表因子及物候導致影像光譜變異對變遷偵測之影響，確認影像同態化對變遷偵測之重 要性，同時評估各影像同態化法之功效。本研究將擬似不變異地物線性迴歸 (PIFLR) 、全影像迴 歸 (IR) 、直方圖匹配(HM) 等三方法應用在台中縣烏石坑地區之十三期 SPOT 影像上，以評估其 同態化效果。就影像外觀、操作難易、均方根誤差及變遷偵測正確率來進行比較，HM 之效果最 優，IR 居次，而 PIFLR 最遜。就波段來看，兩可見光波段之效果優於規整差植生指標 (NDVI) ， 且紅光略優於綠光，而近紅外光最差。變遷偵測正確率降低之主因在無變遷誤授為疑似或有變 遷，HM 及 IR 法能有效改善，PIFLR 法則起伏不定。另一方面，有變遷之漏授雖佔甚小比例卻是 關鍵。各方法之改善漏授效果缺乏規律性，似因缺乏不同試區且具充足之真變遷樣本，仍需深入 研究。然影像同態化對變遷偵測至為重要。 【關鍵字】相對輻射糾正、擬似不變異地物線性迴歸、影像迴歸、直方圖匹配、規整差植生指標 Research paper Radiometric Normalization of Multi-date Satellite Images for Change Detection in the Wu-Shyr-Keng Area of Taichung Prefecture Li-Huey Pan1 Nan-Chang Lo2 Kai-Yi Huang3 Abstract Change detection from multi-date images has been commonly used in monitoring changes 【 】 in landscape. However, spurious changes caused by variations of non-surface factors including illumination, viewing geometry, sensor calibration, and atmospheric effects, as well as surface factors 1.國立中興大學森林學系碩士 Master, Department of Forestry, National Chung-Hsing University. 2.國立中興大學 實驗林管理處技士 Specialist, Experimental Forest Office, National Chung-Hsing University. 3. ( ) 國立中興大學森林學系 教授 通訊作者 Professor, Department of Forestry, National Chung-Hsing University (Corresponding author). 30 多期衛星影像相對輻射糾正於台中縣烏石坑地區變遷偵測之探討 including seasonal phonological differences and topographic effects, may lead to inaccurate results, thereby reducing the accuracy of landscape change detection. The study was intended to examine the influence of spectral varia