The accurate classification approaches for remote sensing image based on GIS and RS

This study selected aquatic vegetation in BaiYangDian Wetland Nature Reserve in Hebei Province. Measuring the spectral reflectance of typical aquatic vegetation included stonewort, reed, lotus and typha, involved high-resolution GeoEye-1(GE) remote sensing image, based on spectral reflectance of different species simulating bands of GE, calculated the simulated bands, educed better bands or band combinations for distinction species; used object-oriented classification method to extract vegetation information in GE image. The spectrum analysis results show that, band 2 to band 4 (BD2/4) is the best combination model of spectra extracted for lotus. BD1/4 is best for stonewort, and there are high distinction at BD4/3 for reed and typha. Classification accuracy of test results show that, the overall accuracy of using the band combination to extract information arrived at 94.4%. This study can provide technical support for monitoring plant of the remote sensing interpretation and vegetation dynamic distribution.