A space-spectrum combination classification method for hyperspectral images based on the four-color theorem

Abstract

The invention discloses a hyperspectral image space-spectrum combination classification method based on the four-color theorem, using SVM to directly obtain the initial rough classification result in the spectral domain of the hyperspectral image, and then utilizing the four-color theorem to mine the spatial information of the hyperspectral image to obtain classification Probability map matrix s; after segmenting and edge extraction of hyperspectral image, use the four-color theorem to mine the spatial information of hyperspectral image to obtain classification probability map matrix r and t; classify probability map matrix r, s and t by element Multiply to get the final classification probability map matrix p, and perform maximum likelihood category judgment on the classification probability map matrix p to get the final classification result. The invention utilizes the four-color theorem to excavate the spatial domain information of the hyperspectral image, and combines it with the spectral domain information to propose a hyperspectral image classification method combining space and spectrum. This method is simple and efficient, and greatly improves the hyperspectral image classification accuracy.

Description

technical field [0001] The invention relates to the fields of remote sensing information processing and intelligent signal processing, in particular to a hyperspectral image space-spectrum combination classification method based on the four-color theorem. Background technique [0002] Hyperspectral remote sensing sensors can acquire hundreds of digital images in continuous bands from visible light to infrared. The energy of electromagnetic waves reflected by different substances is also different, which is manifested in the presence of a unique peak point at a specific wavelength. A large number of digital images of continuous bands make the classification, representation and description of surface objects more accurate and robust. Classifying hyperspectral images is one of the most popular analytical processing methods. In the past, only the information of different bands was used for classification, but the spatial information of image pixels was not fully utilized. In ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the deficiencies in the prior art, the present invention provides a hyperspectral image space-spectrum combination classification method based on the four-color theorem, which solves the problem of inaccurate hyperspectral image classification

Images