A Hyperspectral Mixed Pixel Classification Method Based on Neighbor Cooperative Enhancement

Abstract

The invention provides a hyperspectral mixed pixel classification method based on neighbor synergistic enhancement, comprising: calculating the spectral signature matrix of multi-target features using marked sample features; designing a multi-category classifier based on spectral features to classify the features Classification; integrate the spatial structure features in the classification results, and extract the neighboring pixels; use the neighboring pixels to collaboratively classify the unmarked hyperspectral objects; use the iterative method to gradually classify and label the unmarked objects; The method of domain extension further fuses the spatial characteristics of the target object to complete the final classification and labeling. The invention uses a multi-category classifier to simultaneously classify the object categories, which solves the problem that the traditional classification method cannot classify the background objects; and uses the method of neighbor synergistic enhancement to gradually mark the unmarked object objects, effectively It combines the spectral characteristics and spatial characteristics of ground objects, and the classification effect is better.

Description

technical field [0001] The technical field of hyperspectral image classification of the present invention, in particular, relates to a hyperspectral mixed pixel classification method based on neighbor collaborative enhancement. Background technique [0002] Hyperspectral image classification is an important application in hyperspectral image processing, and its ultimate goal is to classify each pixel in the image. Hyperspectral remote sensing technology uses more spectral bands to make it have a huge advantage in the classification of ground object categories, but the accuracy of ground object spectral information also makes interference and background parts have a certain influence in hyperspectral classification; on the other hand , because the hyperspectral data has the characteristics of large amount of high-dimensional data and small training samples, the Hughes phenomenon is easy to occur during classification. [0003] In recent years, attention has been paid to the ...

AI Technical Summary

Problems solved by technology

At present, most applications use support vector machines and other methods for one-to-one or one-to-many classification; one of the biggest problems with this type of method is that it cannot classify the background part of the hyperspectral image, so when evaluating classification methods, usually They all use the method of removing the background, which leads to the lack of accuracy of this type of method; in addition, this method almost uses pure elements for classification, which leads to the limitation of the classification method and lack of versatility

Images