整合SAR及INSAR資料與技術於地物分類及災害監測; Integration of Sar and Insar Data and Techniques for Landcover Classification and Disaster Monitoring

NCUIR > Research Center > Center for Space and Remote Sensing Research > Research Project > Item 987654321/40057

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/40057

Title:	整合SAR及INSAR資料與技術於地物分類及災害監測;Integration of Sar and Insar Data and Techniques for Landcover Classification and Disaster Monitoring
Authors:	陳錕山
Contributors:	太空及遙測研究中心
Keywords:	太空科技;防災工程
Date:	2001-12-01
Issue Date:	2010-11-03 11:42:32 (UTC+8)
Publisher:	行政院農業委員會
Abstract:	計畫目標: 本計畫擬利用SAR振幅影像( amplitude )、INSAR之同調影像( coherence image )及相位差( phase difference )進行地表分類, 區分地表覆蓋物, 並利用INSAR之同調影像進行變遷偵測以評估災害範圍.另外, 亦利用地物分類結果及災害範圍評估結果, 標定受災之農作區.重要工作項目: 本計畫將利用ERS-1/2及RADARSAT-1之資料進行, 其中包括了SAR之振幅影像、INSAR之同調影像及相位差影像來進行地表覆蓋物分類.分類的方法是利用類神經網路分類器進行監督之地表分類.訓練資料的選擇則是根據實地調查的資料而來.神經網路的輸入為RADARSAT-1之HH或ERS-1/2之VV偏極SAR影像、INSAR之同調影像及相位差影像, 而輸出為地表覆蓋物類別.實行的步驟首先為利用訓練資料訓練神經網路, 俟其完成訓練再令其進行地表覆蓋物分類.另外於災害範圍監測方面, 則利用災害前後所得之兩組SAR資料進行.利用INSAR技術所得到之coherence影像偵測變遷區域, 並於計算變遷區域之強度差及比值以判別受災區域.最後將地表分類之結果、標定之受災面積及地形地物之變化情形, 納入地理資訊系統管理, 以評估災害對於農作區域所造成之受災範圍.預期效益: 本計畫之研究可建立以多時單偏極合成孔徑雷達及干涉合成孔徑雷達來進行地物分類之能力, 亦可以干涉同調影像進行變遷之偵測, 同時整合合成孔徑雷達、干涉合成孔徑雷達及差分干涉合成孔徑雷達於災害監測之能力, 最後評估以干涉合成孔徑雷達計算淹水深度之可行性. This project performs landcover classification using amplitude image of SAR, coherence image and phase-difference of INSAR.It also evaluates the area of disaster including damaged agricultural area from results of landcover classification.The remotely sensed data used in landcover classification are ERS-1/2and RADARSAT-1data by computing the amplitude image, coherence image, and phase-difference image from a serie of multi-temporal SAR.A supervised neural classifier is applied in this purpose.The training data is selected according to the ground-truth from the field trip and assistance of base maps.At training stage, the input channels of neural classifier are the amplitude, coherence, and phase-difference data of training set; the output is the desired class.After training, the classifier performs classification of the whole data according to the well-trained neural network weighting.On the other hand, the evaluation of damaged area is performed via change-detection using coherence image.Meanwhile, the difference and ratio of SAR amplitude before-and after-disaster are also calculated for identification of damaged area.Finally, the results of landcover classification and damaged area evaluation are recorded and managed by geographic information system ( GIS ).The project is aimed at developing the ability of landcover classification by integration of multi-temporal single-polarized SAR and interferometric SAR ( INSAR ), also the ability of change-detection using coherence image of INSAR.It also evaluate the ability of detection of flooded-depth using INSAR. 研究期間：9001 ~ 9012
Relation:	財團法人國家實驗研究院科技政策研究與資訊中心
Appears in Collections:	[Center for Space and Remote Sensing Research ] Research Project

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	405	View/Open

社群 sharing

Loading...