Research On Hyperspectral Images Sparese Unmixing Based Onbionic Intelligent Optimizational Gorithm

Hyperspectral imaging technology uses an imaging spectrometer to image the same scene in different bands respectively.The imaging spectrometer is capable of receiving a wide range of wavelengths,a large number of electromagnetic bands,and a small band width,so that a large number of mixed pixels exist.Mixed pixels are made up of a mixture of pure substances.In order to achieve accurate classification of the ground objects,it is necessary to decompose the mixed pixels of the image,which is the hyperspectral image unmixing.Hyperspectral image unmixing has been widely used in various fields such as military,domain,mecicine,and environmental protection.Therefore,it is of great significance to study the method of unmixing.The semi-supervised sparse unmixing method is to reconstruct the hyperspectral image from the spectral library by selecting the end set under the premise of the known spectral library.Since the method has certain prior knowledge,the hypothetical pure pixel is avoided.The sparse unmixing methods based on spectral libraries have attracted much attention.In this paper,the advantages of bionic intelligent optimization algorithm are analyzed and the whale group optimization algorithm is improved.The problem of unmixing under the sparse model framework is studied and its defects are found.The bionic intelligent optimization algorithm is combined with the hyperspectral sparse unmixing to further improve the unmixing precision.The main work of the thesis is as follows:(1)In order to improve its convergence accuracy and speed up the convergence,a whale group optimization algorithm based on nonlinear control factor and mutation strategy.is proposed in this paper.Firstly,the linear control factor in the original algorithm is replaced by a nonlinear control factor containing the tangent function,which controls the search direction of the overflow boundary search agent to expand the search range and speed up the convergence.Secondly,a new mutation strategy is added to avoid the population iterating.around the current minimum,enhancing the global search capabilities.Simulation experiments show that the whale group optimization algorithm based on nonlinear control factor and mutation strategy has improved convergence accuracy and convergence speed,and shows better optimization performance.(2)As the unmixing accuracy of the subspace matching tracking algorithm in thegreedy method is low and the sparsity is not strong,combining the improved whale algorithm with sparse unmixing,this paper proposes a sparse unmixing using the improved whale optimization subspace matching pursuit algorithm.Based on the subspace matching pursuit algorithm,the constrained sparse regression is used as the objective function,and the improved whale group optimization algorithm is used to solve the abundance coefficient of the known endmember set.With a standard of the minimum reconstruction error,the redundant endmembers with small coefficients are removed to the greatest extent,so as to improve the accuracy of endmembers extraction of the subspace matching pursuit algorithm,and further improve the unmixing accuracy of hyperspectral images.The synthetic image experiment and real remote sensing image experiments show that the proposed algorithm effectively removes a large number of redundant end elements and has higher unmixing precision and sparsity.