Visualization method capable of sparsely representing hyperspectral image

Abstract

The invention belongs to the technical field of a hyperspectral remote sensing image and in particular relates to a visualization method capable of sparsely representing a hyperspectral image. The method comprises the following steps: (1) determining color tags of dictionary atoms; (2) performing color mixing according to the colors of the selected dictionary atoms and sparse coefficients of pixels to generate mixed color images; and (3) displaying the coefficient of each pixel to form a detailed drawing. The method can follow a distance preservation property and has larger separability among classes so as to guarantee that the generated image is significant in color change and has better visual effect; surface objects can be clearly and macroscopically displayed, the specific constitution situation of each picture element can be intuitively and completely displayed, and the original hyperspectral image (HSI) is reconstructed according to the detailed information.

Description

technical field [0001] The invention belongs to the technical field of hyperspectral remote sensing image visualization, and in particular relates to a visualization method for sparse representation of hyperspectral images. Background technique [0002] Since the advent of imaging spectrometers in the 1980s, hyperspectral images (Hyperspectral imagery, HSI) have been more and more widely used in military, agriculture, ocean, environment, etc., and the corresponding visualization technology has also attracted increasing attention from researchers at home and abroad. Hyperspectral data covers a wide range, and contains both ground object space and spectral information. The relative area of ​​the target is small, the background is more complex, and it is prone to noise, blur, shadows, etc. In addition, there are a large number of mixed pixels in HSI, and the analysis, processing and display of mixed pixels have become an important obstacle to the further development of remote s...

AI Technical Summary

Problems solved by technology

A single image of sparse coefficients is too independent to analyze the results of sparse coefficients as a whole

The two-dimensional coefficient change image can only display the change of one pixel vector, and the original data pixels use more pixels in the generated image, so it is not suitable to display more pixels

In addition, both methods lose the spatial characteristics of the image

The method of only displaying the classification results is not intuitive, the spectral information is seriously lost, and the visualization effect depends on the classification performance of the subsequent method

[0005] Q.Du, S.Cai, M.Cui et al proposed several display methods of hyperspectral mixed pixels, but these methods still cannot meet the display requirements of HSI with high information content

Although the multi-layer display method mentioned in it can display more information in the same image, this method still cannot fully express the spectral information of HSI

And these display methods are displayed under certain assumptions, that is, the mixed pixel is a linear mixture of end members, and the mixed spectrum is not completely linear, so the above methods cannot correctly and intuitively express the spectral information of the mixed pixel

In addition, the existing HSI visualization methods are unidirectional and cannot reconstruct the original hyperspectral data cube through image data, which will lead to a decrease in information utilization

Images