Hyperspectral Band Selection Method Based on Normalized Multidimensional Mutual Information and Clonal Selection

Abstract

The invention relates to a dimensionality reduction method for a hyperspectral image, in particular to a hyperspectral band selection algorithm based on normalized multidimensional mutual information and clone selection. A method for selecting the number of cloning iterations in hyperspectral image band selection is provided. The steps of the present invention are as follows: 1. Read in the hyperspectral image, define the antigen and randomly generate an initial set, and select the best individual composition set according to the individual fitness value. 2. Clone the best individual to generate a temporary clone set, perform high-frequency mutation operation on the clone set, and select the best individual to form a set again. 3. Use the normalized multidimensional mutual information to judge the correlation degree of the two sets before and after to decide whether to stop the iteration. The invention can achieve the purpose of reducing the dimension of hyperspectral images. In order to make the numerical calculation more accurate, the number of iterations is selected by using the normalized multi-dimensional mutual information, which reduces the unnecessary excessive iteration process in the selection process, and is suitable for hyperspectral images. Image band selection application.

Description

(1) Technical field [0001] The invention relates to a dimensionality reduction method for a hyperspectral image, in particular to a hyperspectral band selection method based on normalized multidimensional mutual information and clone selection. (2) Background technology [0002] Hyperspectral imagery is a mass data source that combines imagery and spectrum. It contains both image information and spectral information. It can not only give spectral intensity data of each pixel on each spectral segment, but also has a high spectral resolution. This imaging technology can be applied in the field of target recognition, providing a good detection method for the detection and search of airborne hyperspectral imagers. However, hyperspectral images contain too much information and there is redundant information, so it is necessary to perform feature selection on hyperspectral data. [0003] The main task of feature selection is to select the features that can represent the original ...

AI Technical Summary

Problems solved by technology

It can solve the difficult problem of directly solving the multidimensional mutual information and the problem that the number of iterations needs to be determined manually

Images