abstract
Hyperspectral remote sensing images with dozens of to hundreds of each band of data, for people to understand the features provides abundant information, the classification and features of recognition is very favorable targets. Hyperspectral remote sensing data recorded the continuous spectrum of terrain target. And other types of remote sensing image, more information than rich, consequently have recognition more kinds of geophysics target with higher precision, and the ability of target classification. But hyperspectral remote sensing images, because of large amount of data, and there is a data redundancy and is currently in the phenomenon of its classification, often need first to data dimension reduction, reuse some other classification algorithm classification. To do so, the increased workload sorting failed to achieve "integration", "automatic".
The decision tree algorithm has flexible, intuitive, clear, strong, operational efficiency higher characteristic, in remote sensing image classification field is worth studying. At present more mature decision tree algorithm has ID3 algorithm, C4.5 algorithm, CART algorithm, C5.0 algorithm, etc, they are suitable for discrete data sets the classification, and processing data sets size is limited, so used to deal with remote sensing data is there are some deficiencies, need to study a suitable for remote sensing image classification decision tree.