摘要
Using three-phase remote sensing images of China-Brazil Earth Resources Satellite 02B (CBERS02B) and Landsat-5 TM, tobacco field was extracted by the analysis of time series image based on the different phenological characteristics between tobacco and other crops. The spectral characteristics of tobacco and corn in luxuriant growth stage are very similar, which makes them difficult to be distinguished using a single-phase remote sensing image. Field film after tobacco seedlings transplanting can be used as significant sign to identify tobacco. Remote sensing interpre- tation map based on the fusion image of TM and CBERS02B's High-Resolution (HR) camera image was used as stan- dard reference material to evaluate the classification accuracy of Spectral Angle Mapper (SAM) and Maximum Like- lihood Classifier (MLC) for time series image based on full samples test method. SAM has higher classification accu- racy and stability than MLC in dealing with time series image. The accuracy and Kappa of tobacco coverage extracted by SAM are 83.4% and 0.692 respectively, which can achieve the accuracy required by tobacco coverage measurement in a large area.
Using three-phase remote sensing images of China-Brazil Earth Resources Satellite 02B （CBERS02B） and Landsat-5 TM, tobacco field was extracted by the analysis of time series image based on the different phenological characteristics between tobacco and other crops. The spectral characteristics of tobacco and corn in luxuriant growth stage are very similar, which makes them difficult to be distinguished using a single-phase remote sensing image. Field film after tobacco seedlings transplanting can be used as significant sign to identify tobacco. Remote sensing interpretation map based on the fusion image of TM and CBERS02B＇s High-Resolution （HR） camera image was used as standard reference material to evaluate the classification accuracy of Spectral Angle Mapper （SAM） and Maximum Likelihood Classifier （MLC） for time series image based on full samples test method. SAM has higher classification accuracy and stability than MLC in dealing with time series image. The accuracy and Kappa of tobacco coverage extracted by SAM are 83.4% and 0.692 respectively, which can achieve the accuracy required by tobacco coverage measurement in a large area.
基金
Under the auspices of China Postdoctoral Science Foundation (No. 20080430586, 20070420018)
National Natural Science Foundation of China (No. 40801161, 40801172)
Sino US International Cooperation in Science and Technology (No. 2007DFA20640)
