Hyperspectral image classification method based on stratified probability model

Abstract

The invention discloses a hyperspectral image classification method based on a stratified probability model, and belongs to the technical field of image processing. The hyperspectral image classification process comprises the steps that hyperspectral images are preprocessed to obtain feature image data; data labeled by an expert label are picked out from the feature image data and used for training the stratified probability model; training data and test data are separated by the feature image data according to the expert label and projected to the trained stratified probability model, and new feature representation of the training data and the test data is obtained; a classification result is obtained by conducting supervised classification through a support vector machine. Compared with the existing mode of directly conducting support vector machine classification on original hyperspectral data, the hyperspectral image classification method has the advantages that higher in classification accuracy and more homogeneous classification regions are obtained, and can be applied to hyperspectral image classification.

Description

technical field [0001] The invention belongs to the field of image processing and relates to a hyperspectral image classification method. Background technique [0002] The development of hyperspectral remote sensing images benefits from the development and maturity of imaging spectroscopy technology. Through hyperspectral sensors mounted on different space platforms, that is, imaging spectrometers, in the ultraviolet, visible, near-infrared and mid-infrared regions of the electromagnetic spectrum, the target area is simultaneously imaged in tens to hundreds of continuous and subdivided spectral bands . While obtaining surface image information, it also obtains its spectral information, achieving the combination of spectrum and image. At present, remote sensing systems with a wavelength interval of less than 10nm and more than 36 bands are generally defined as hyperspectral remote sensing. [0003] The biggest feature of hyperspectral images is the combination of imaging t...

AI Technical Summary

Problems solved by technology

Some time ago, hyperspectral image object classification mainly focused on relying only on the feature information of the spectral domain for classification, but this lost the image domain, that is, the information of the graphic structure or adjacent relationship in the spatial domain.

This does not make full use of the information provided by hyperspectral imagery

Images