Hyperspectral image sparse unmixing device based on compressed sensing and method for unmixing hyperspectral image

Abstract

The invention provides a hyperspectral image sparse unmixing device based on compressed sensing. A sampling box and a host are installed on a workbench; the bottom of a loading drawer is in contact with the top of a heat exchange plate; a plurality of ventilation heat pipes are arranged at the bottom of the heat exchange plate; and a filter lens and an LED lamp tube correspond to a sample located on the loading drawer. Meanwhile, the invention provides a method for unmixing a hyperspectral image by using the hyperspectral image sparse unmixing device based on compressed sensing. A sample is placed on the loading drawer, and normal-temperature air enters the ventilation heat pipes through an air supply chamber via a gas shunting pipe to absorb redundant heat; and the sample is compressed and collected through the filter lens, and a complete hyperspectral signal of the sample is reconstructed by a matching pursuit algorithm. According to the invention, the filter lens and a single-pixel camera are additionally arranged on the sampling box, sparse unmixing is carried out by a Gaussian random measurement matrix and a matching pursuit algorithm, and the heat exchange plate, the ventilation heat pipes, the gas shunting pipe and a gas collecting pipe are arranged, so that a heat dissipation function is achieved.

Description

technical field [0001] The invention relates to the technical field of hyperspectral image processing, in particular to a hyperspectral image sparse unmixing device and a hyperspectral image unmixing method based on compressed sensing. Background technique [0002] Hyperspectral images have high spectral resolution, and each pixel simultaneously records signal values ​​of tens to hundreds of bands, but are limited by relatively low spatial resolution and the natural mixture of different ground objects. Hyperspectral images Mixed pixels that are superimposed by the spectra of multiple substances are ubiquitous; and the measurement spectrum of mixed pixels is decomposed into endmember spectra of pure substances, and the abundance ratio of each endmember is estimated, which is the so-called spectral solution. Sparse unmixing is the mainstream semi-supervised unmixing method in recent years. It uses the complete spectral library as the candidate set of endmembers, and integrates...

AI Technical Summary

Problems solved by technology

[0003] However, most of the traditional hyperspectral image sparse unmixing devices have the following problems: First, the scale of the spectral library as a set of endmember candidates in reality is increasing day by day, and the number of candidate endmembers is much larger than the number of measured endmembers, which makes the performance of a large number of endmembers Redundant characteristics lead to problems such as high cost, data storage or transmission, and difficulty in obtaining effective information during sparse unmixing; second, there is no heat dissipation function, and the sample is heated by light during the imaging process, which affects the integrity of the sample information characteristics , reduce the accuracy of unmixing, and at the same time, light and heat also increase the air pressure inside the device, damage the parts of the device, and shorten the service life of the device

Images