Automated morphological endmember extraction based hyperspectral image data unmixing method

Abstract

The invention relates to an automated morphological endmember extraction based hyperspectral image data unmixing method and belongs to the technical field of image data processing. According to the method, calculation is performed by utilizing a reference spectral vector to obtain an image element with highest mixing degree in a whole image, and based on this, a morphological operator is improved; an original structural element is improved with a new MEI calculation method; a structural element of an odd size is replaced with a structural element of an even size; and four candidate endmembers are selected from each structural element, so that possible information loss is effectively avoided, the accuracy of data processing is further improved, and finally the definition of the image obtained after data processing is improved; and the automated morphological endmember extraction based hyperspectral image data unmixing method is quite simple and convenient to implement.

Description

technical field [0001] The present invention relates to the technical field of image data processing, in particular to the technical field of hyperspectral image data processing, in particular to a hyperspectral image data unmixing method based on automatic morphological endmember extraction. Background technique [0002] Hyperspectral images have higher spectral resolution and can provide more detailed spectral information of ground objects, so they can be better used for remote sensing image classification and target recognition. At the same time, due to the low spatial resolution of hyperspectral sensors, mixed pixels are prevalent in hyperspectral images. Endmember extraction and spectral unmixing are the two core tasks of mixed pixel linear decomposition, and endmember extraction is the premise of mixed pixel decomposition. Typical endmember extraction algorithms can be divided into three types: methods based on projection, such as Pure Pixel Index (PurePixelIndex, PPI...

AI Technical Summary

Problems solved by technology

First of all, when AMEE algorithm defines the morphological operator, the average value of all the pixels in the structure element is regarded as the most mixed pixel. However, when the structure element contains more pure pixels, the average value of the pixels is closer to the pure pixels, then the dilation and erosion operations will not get the expected results

Secondly, the original Morphological Eccentricity Index (MorphologicalEccentricityIndex, MEI) indicates the spectral angular distance between the purest pixel and the most mixed pixel in the structural element, and the most mixed pixel in different structural elements is generally different , so the absolute size of the MEI cannot really represent the purity of the pixel

Finally, in the algorithm, only one pixel is selected from each structural element as a candidate end member. When the structural element contains two or more pure pixels of different substances, important pixels may be lost.

Images