A feature selection method based on memory multipoint cross-gravity search

Abstract

The invention discloses a feature selection method based on memory multi-point cross-gravity search. The method sets each particle as an alternative solution of an optimal feature subset, and evaluates the quality of the alternative solution through the band subset evaluation function , and guide the particles to exchange information to complete rapid convergence; the solution corresponding to the particles with the best quality is the optimal spectral feature subset. In order to improve the adaptability of the algorithm, the present invention proposes an information exchange mechanism based on the degree of population evolution based on the gravitational search algorithm: in the exploration stage, based on the multi-point crossover strategy, it can fully learn from other particles in the population and conduct extensive searches; Learn from your own optimal experience to ensure rapid convergence. The present invention can select an optimal spectral feature subset with a small number of bands and can obtain stable classification results, thereby solving the problems of complex calculation and low classification accuracy caused by high redundancy of hyperspectral remote sensing image data.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, in particular to a feature selection method based on memory multi-point cross-gravity search. Background technique [0002] With the development of sensor technology, hyperspectral remote sensing images began to play an important role in environmental monitoring, land and resources survey and evaluation, urban planning and other fields after the 1980s. In this process, the classification accuracy of remote sensing images directly determines the effectiveness of its application. Although the rich spectral features of hyperspectral remote sensing images can more accurately describe the ground cover information, such massive spectral information usually has high redundancy, which will cause the Hughes phenomenon in the classification process and consume a lot of storage and computing space. Therefore, how to identify the most effective small number of bands from the high-di...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the demand for remote sensing image feature selection and the shortcomings of the gravitational search algorithm, the purpose of the present invention is to provide a feature selection method based on memory multi-point cross-gravity search, through self-adaptive balance exploration and development optimization feature subset evaluation function , and mutate the global historical optimum to effectively improve the problems of insufficient development ability, lack of memory, and premature convergence of the gravitational search algorithm, and finally obtain the optimal spectral feature subset with a small number of bands and stable classification results

Images