Extraction of diseased pine in Qinling Mountains based on hyperspectral data
This paper proposes an integrated classification method for hyperspectral images that optimizes subspace partitioning to address the problem of poor extraction efficiency of Qinling disease pine caused by the large amount of hyperspectral image data and severe information redundancy.This method uses mean square error and Jeffreys-Matusita(J-M)distance to optimize subspace partitioning algorithms to select the optimal band in hyperspectral images.By inputting the optimal band into an ensemble classifier composed of artificial neural networks,support vector machines,and maximum likelihood methods,the diseased pine patches in forest land are extracted.The experimental results show that the ensemble classi-fication of 10 optimal band combinations can reduce the data dimension while maintaining the spectral features of the image,with a total classification accuracy of 97.75％.It can not only select the bands that play a key role in extracting disease pine,but also make up for the shortcomings of a single classifier.
NETL
NSTL
万方数据
维普
