Small sample hyperspectral image classification method based on supervised self-contrast learning

Abstract

The invention discloses a supervised self-contrast learning-based small sample hyperspectral image classification method. The method comprises the following steps of obtaining a hyperspectral image to be classified; the to-be-classified hyperspectral images are input into the supervised coding model C1, feature vectors of the to-be-classified hyperspectral images are obtained, the feature vectors of the to-be-classified hyperspectral images are input into the classifier C2, the classification result of the classified hyperspectral images is obtained, and classification of the hyperspectral images can be achieved through fewer samples.

Description

technical field [0001] The invention belongs to the field of remote sensing image processing, and relates to a small-sample hyperspectral image classification method based on supervised self-comparison learning. Background technique [0002] Hyperspectral images are usually special images taken by drones and other aircraft with hyperspectral imaging devices, which contain more bands and higher resolution than ordinary images, and can be continuously imaged in a certain band, including A large amount of spatial information and spectral information of ground objects is widely used in the field of earth observation, and plays an important role in economic, agricultural, and environmental monitoring. [0003] Hyperspectral image classification refers to distinguishing each pixel in the image according to the obtained sample characteristics, and classifying the category to which it belongs. Hyperspectral image classification methods in the field of image processing mainly rely o...

AI Technical Summary

Problems solved by technology

However, in the actual classification of hyperspectral images, due to the high-dimensional features of hyperspectral images, there will be Hughes phenomenon, that is, in the case of direct image classification in the case of limited samples, the classification effect will decrease with the increase of dimension, such as volume Although the product neural network method has excellent performance, it often requires a very large amount of labeled training data to train an excellent classifier, and when there are few training samples, it is very prone to overfitting problems and the performance is relatively poor

Images