Hyperspectral image-oriented classification method and system

Abstract

The invention discloses a hyperspectral image-oriented classification method and a hyperspectral image-oriented classification system. The hyperspectral image-oriented classification method comprises the following steps: extracting spatial and spectral features of a hyperspectral image by adopting 3D-DWT (three-dimensional discrete wavelet transform); dimensionality reduction is carried out on a new hyperspectral image obtained after the spatial and spectral features of the hyperspectral image are extracted; segmenting the hyperspectral image after dimension reduction into overlapped small image blocks; and classifying the small image blocks by using a pre-trained convolutional neural network classifier. According to the method, the 3D-DWT and the CNN are combined, and the 3D-DWT is used as a preprocessing means of the hyperspectral image, so that the robust feature expression capability is obtained, the calculation burden of the CNN is reduced, and the higher classification performance can be obtained under the condition of fewer training samples; the problem that hyperspectral data lacks training samples with labels is greatly relieved.

Description

technical field [0001] The invention relates to an image processing technology of remote sensing images (hyperspectral images), in particular to a classification method and system for hyperspectral images. Background technique [0002] At present, most of the deep learning methods are used for the classification of hyperspectral images, and the Convolutional Neural Network (CNN) is the most widely used due to its better performance and unique structure. However, most of the classification structures based on CNN belong to the category of supervised learning, which leads to a large number of parameters to be trained in the model, which means that a large number of labeled samples are needed to ensure the classification performance of the algorithm. Since the cost of obtaining labeled hyperspectral data is extremely high, the number of labeled samples is very limited. At the same time, many CNN-based classification algorithms often adopt extremely complex classification struc...

AI Technical Summary

Problems solved by technology

However, the joint use of 3-D DWT and CNN is still very rare.

Images