ABSTRACT
Hyperspectral image has a number of multi-band, high spectral resolutions, narrow bandwidth the huge amount of data and other characteristics, and the strong correlation between their bands, higher data redundancy, this for hyperspectral image classification and recognition has brought great diculties. Band selection for hyperspectral image can be reduced dimensionality of hyperspectral image data on the basis of possible reserves object recognition rate, and reduces the computational complexity of the image classification, improve the precision of classification, which is currently an important direction of processing hyperspectral dimensionality reduction study [1]. At present more mature traditional band selection algorithm is divided into two categories:
1 Optimal band selection based on the amount of information, such as entropy and joint entropy, covariance matrix eigenvalues, optimal index factor, adaptive band selection and other methods.
