| Abstraction of the abundance information of broadleaf forest, coniferous forestand scrubby vegetable from TM remote sensing images has great importance to forestresource survey and ecology construction.Extracting the abundance information ofthree different types of forest with linear mixed model is a remote sensing softclassification method and the method is more scientific accurate than remote sensinghard classification. The abundance information of broadleaf forest, coniferous forestand scrubby vegetable from TM remote sensing images of Meijiang Basin had beenextracted with linear mixed model.To improve the accuracy of extraction of thevegetation abundance information with linear mixed model,in this theses, not onlysome general methods that include Minimum Noise Fraction(MNF),Pixel PurityIndex (PPI) and n-D Visualizer had been adopted, but also three other data processinglinks that include topographic correction, masking some components using suitableNDVI threshold value and reflectivity normalization had been added.The results showed that three other data processing links added can improve theaccuracy of extraction of the vegetation abundance information, and also showed thatthe most value of RMS(root mean square error) was less than0.026, the maximumvalue of RMS was0.049,the minimum value of RMS was-0.024. They indicated theimmixing of Linear Mixed Model was successful. The60sample points which wererandomly sampled and uniformly distributed in the field survive were used in linearfitting test of the simulation results. The correlation coefficient R2between thesimulation value of broad-leaved forest abundance and the measured value ofbroad-leaved forest abundance was0.8358.The correlation coefficient R2between thesimulation value of coniferous forest abundance and the measured value of coniferousforest abundance was0.8861.The correlation coefficient R2between the simulationvalue of low vegetation abundance and the measured value of low vegetationabundance was0.8202. |
PDF Full Text Request