Study Of Hyperspectral Remote Sensing Image Endmember Extraction Based On Multi-objective Optimization

Since the development of hyperspectral remote sensing technology,it has been developing rapidly in the past 30 years.Hyperspectral remote sensing has become an important means of surface observation,and the surface images obtained by hyperspectral imager contain a large amount of spectral,radiometric and spatial information.Hyperspectral remote sensing image data has been widely used in resource monitoring,environmental monitoring,precision agriculture,land surveying and mapping,military reconnaissance and so on.With the development of hyperspectral remote sensing technology,hyperspectral remote sensing image processing and information extraction technology is facing severe challenges.The spectral resolution of hyperspectral remote sensing image is nanometer,and the spatial resolution of the sensor is relatively low,coupled with the earth's surface complex environment and uneven distribution,will lead to the hyperspectral remote sensing image mixed pixel is widespread.Presence of mixed pixels Seriously affect the classification accuracy and the recognition of ground features.How to decompose the mixed pixels accurately and rapidly has become one of the key problems in hyperspectral remote sensing.The influence of hyperspectral remote sensing spectral unmixing of the two basic steps,determine the types of mixed pixels in the extraction and calculation of the various basic features that the proportion is called abundance inversion in the mixed pixel for endmember.In this paper,the endmember extraction method of hyperspectral remote sensing image is studied,and the reliability of the proposed method is verified by comparing the endmember and spectral library.In this paper,the problem of endmember extraction of hyperspectral remote sensing image is studied:1.Briefly describes the hyperspectral remote sensing technology and endmember extraction algorithm development,two mixed model of hyperspectral remote sensing image,the linear spectral mixture model is introduced in detail and the nonlinear spectral mixture model gives a brief introduction of linear spectral mixture model is introduced.Several classic yuan the extraction algorithm based on the summary of the advantages and disadvantages of these algorithms.2.Aiming at the defect extraction algorithm is based on several endmember linear mixed model,we focus on two endmember extraction algorithms based on combinatorial optimization of hyperspectral remote sensing images.The discretization method and the design of operator are also summarized.Finally,the performance of the algorithm and the reliability of the extracted endmembers are analyzed through experiments3.The hyperspectral remote sensing image endmember extraction problem is modeled as a multi-objective optimization problem is solved,and designed an leader selection strategy can obtain a series of containing different number of endmembers endmember subset through multi-objective optimization algorithm,need not be the same as the third chapter that need to reset the constant number of endmembers.The feasibility of the proposed method is verified by experiments.The proposed algorithm is compared with several widely used endmember extraction algorithms,and the experimental results are analyzed.