Research on Spectral Mapping models and its Accuracy Evaluation Based on Pixel Feature Vectors of Hyperspectral Imagery

摘要

Hyperspectral remote sensing (MRS) imagery classification and mapping is currently a research hotspot of remote sensing technology with image pixel spectral vectors. The paper researched some models of hyperspectral image classification based on spectral angle mapper (SAM), support vector machines (SVMs) and Minimum Noise Fraction (MNF) technology, and structured three HRS image classifiers of MNF_SAM, SVM and MNF_SVM models combined with the AVIR1S hyperspectral image. The research results indicate that the SVM classification accuracy is higher than the MNF_SAM, but MNF_SVM classification efficiency and accuracy is better than the MNF_SAM and SVM classifier.