Hyperspectral image classification method and system based on domain adaptation

Abstract

The invention discloses a hyperspectral image classification method and system based on domain adaptation, and relates to the field of image processing, and the method comprises the steps: carrying out the first processing of first source domain data, and obtaining a first training data set; performing second processing on the first target domain data to obtain a second training data set; performing feature extraction network training based on the training data set to obtain a first feature extraction training result; performing feature classification of the first feature extraction training result to obtain a first feature classification result; and after performing first processing on the first feature extraction training result and the first feature classification result, connecting a domain classification network, and obtaining a domain classification result based on the domain classification network. The technical problems that in the prior art, due to the fact that the number of spectral bands is large, information redundancy is likely to be caused, the classification interpretation process is difficult, parameter optimization of a network model is difficult to carry out by using limited training samples in a deep learning method, and the classification precision is affected are solved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method and system for classifying hyperspectral images based on domain adaptation. Background technique [0002] The hyperspectral image is meticulously segmented in the spectral dimension, and there are also N channels in the spectral dimension. Therefore, what is obtained through the hyperspectral device is a data cube, which not only contains the information of the image, but also expands in the spectral dimension. As a result, not only the spectral data of each point on the image, but also the image information of any spectral band can be obtained. Hyperspectral image classification is of great significance in the fields of precision agriculture, classification and recognition of urban features, and military applications. The existing technology to classify hyperspectral images mainly uses labeled hyperspectral images as training samples, and extracts sample features. Class...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a hyperspectral image classification method and system based on domain adaptation, which solves the problem that the existing hyperspectral image classification is prone to information redundancy due to the large number of spectral bands, and the classification and interpretation process is difficult. In the deep learning method, it is difficult to use limited training samples to optimize the parameters of the network model, thereby affecting the technical problem of classification accuracy. Through the method based on domain adaptation, the domain adaptation and automatic feature learning are combined, and the target domain of unlabeled samples is used. , and the source domain containing labeled samples are trained at the same time to ensure the consistency of cross-domain labeled samples, realize cross-domain sample migration, increase the amount of labeled sample data, provide sufficient sample support for model parameter optimization, and solve network problems with limited labeled samples. The problem of difficult convergence, so as to achieve the technical effect of accurate classification of hyperspectral images

Images