Abstract Based on multi-spectral remote sensing image data of CBERS and using information quantity, correlation
matrix and OIF method, the authors analyzed image band characteristics and obtained the overall cognition of image
data. Through drawing spectral curves of land cover sample average values, this paper deals with land cover
spectral characteristics in five bands and describes land cover reflection characteristics of each band. Based on
typical land cover sample data, the authors drew the box plots for five bands and two means of PC1 and NDVI and
further analyzed land cover characteristics. Using the Z-test method, this paper got the optimal texture measures
for the discrimination of different land covers. Using the object-oriented classification technique, the authors
carried out classification experiments, and the results obtained have verified the effectiveness of applying CBERS
remote sensing imagery to wetland cover classification, enriched methods for inland wetland cover extraction, and
expanded the application field of CBERS remote sensing image.
|