A hyperspectral image classification method based on superpixel sample expansion and generative adversarial network

Abstract

The invention provides a hyperspectral image classification method based on superpixel sample expansion and generative adversarial network, and aims to solve the problem of low classification accuracycaused by network overfitting when the number of labeled training samples is small. The method comprises the following steps: constructing an initial training set and a test set, and performing expansion to obtain an expansion training set and a candidate test set; constructing a generative adversarial network consisting of a generator and a discriminator; using a generator to generate a false sample, using a discriminator to obtain a true and false prediction label and a category prediction label of the false sample and the extended training set; constructing loss functions of the generatorand the discriminator, and alternately training the generator and the discriminator; training a support vector machine; tnabling the candidate test set to pass through a trained discriminator and a trained support vector machine to obtain a candidate tag set; and determining category labels of the test set for the candidate label set by using a maximum voting algorithm. According to the method, the spatial features of the hyperspectral image are effectively extracted, the overfitting problem is relieved, the classification accuracy is improved, and the method can be used for carrying out ground object classification on the hyperspectral image.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to an image classification method, and further relates to a hyperspectral image classification method based on superpixel expansion and generation confrontation network. The invention can be used to classify the hyperspectral images. Background technique [0002] Compared with ordinary sensors, hyperspectral imaging spectrometers have more spectral channels and narrower wavelength ranges in a certain wavelength range. Due to the spectral overlap between bands, hyperspectral images have high spectral resolution. Combined with the spatial information of ground objects acquired by imaging spectrometers, hyperspectral images contain rich one-dimensional spectral information and two-dimensional spatial information at the same time. Hyperspectral image classification technology is one of the hotspots in the field of hyperspectral data applications. To avoid overfitting, the training...

AI Technical Summary

Problems solved by technology

However, the disadvantage of this method is that the number of executions of RGF is not easy to determine, so that the spectral-spatial information cannot be fully combined, and the classification accuracy is low.

Images