Hyperspectral remote sensing image classification method based on attention mechanism and convolution neural network

Abstract

The invention provides a Hyperspectral remote sensing image classification method based on attention mechanism and convolution neural network.featuring that that original hyperspectral remote sensingimage is reduced in dimension by principal component analysis, and the reduce hyperspectral data is sampled into blocks. After that, 3D convolution and pooling operations are carried out to obtain theintermediate feature map. Then, each spectral vector of the intermediate feature is multiplied with the spectral attention module and each spatial feature is multiplied with the spatial attention module to obtain an attention enhancement sample. After that, the convolution operation and attention enhancement operation are performed again. Then the intermediate feature map obtained by 3D convolution operation is inputted into the classifier for classification. The invention has the advantages that the classification cost is reduced, the classification performance is improved, the adaptive feature thinning is realized through the extraction and enhancement of the sample features, and the classification accuracy of the hyperspectral remote sensing image is further improved.

Description

technical field [0001] The invention relates to the field of hyperspectral image classification, in particular to a hyperspectral remote sensing image classification method based on an attention mechanism and a convolutional neural network. Background technique [0002] Remote sensing is a long-distance, non-contact target detection technology and method, and it is an important means for people to study the characteristics of ground objects. With the rapid development of hardware technology and the continuous growth of application requirements, the obtained remote sensing images have gradually developed from wide-band imaging to narrow-band imaging, and at the same time present the characteristics of high spatial resolution, high spectral resolution, and high temporal resolution. Remote sensing was born from this. Hyperspectral remote sensing technology is a very iconic achievement in the history of remote sensing development. Its rapid development has attracted extensive a...

AI Technical Summary

Problems solved by technology

In the deep learning method, the deep neural network is the representative. The convolutional neural network in the deep neural network has achieved good applications in the classification of hyperspectral remote sensing images. However, the amount of input information of the convolutional neural network is not completely positively correlated with the classification effect. , under a certain model, too complex input will not only increase the training time and classification time, but even cause the accuracy to decrease instead of increase.

Images