Research On Dimensionality Reduction Method Of Hyperspectral Remote Sensing Image Based On Whale Optimization Algorithm

Hyperspectral remote sensing image covers rich features of ground object information due to its own high dimensional characteristics.However,the spectral features of adjacent bands of hyperspectral remote sensing images are strongly correlated,so the information redundancy is greatly increased,which increases the difficulty of image processing and classification calculation.Therefore,it is an important content of hyperspectral data processing how to reduce the amount of data and the dimension of data under the premise of ensuring data integrity as much as possible.The high-dimensional data characteristics of hyperspectral remote sensing images lead to dimensionality disaster easily.Therefore,it is necessary to reduce the dimensionality of the original data.Feature selection,as one of the important methods of dimension reduction,has been widely used in dimension reduction.Therefore,how to select effective bands from dozens or even hundreds of bands for subsequent processing and extract effective information from them has become a key problem in dimension-reduction processing of hyperspectral remote sensing data.Whale Optimization Algorithm(WOA)is a new kind of swarm intelligence Algorithm inspired by the strategy of humpback whales in the process of predation.It carries out random search in the whole data space through the constant change of search agents.In the original space,the bands generated by band selection are basically concentrated in a continuous subspace,and they are often very similar,which is easy to cause information reuse and affect the effect of subsequent processing.In order to solve these problems,this paper uses the method of subspace division of wave segment,which divides all bands of hyperspectral data set into several subspaces,and then makes band selection.Ordinary smoothing de-noising methods sometimes cross the boundary of ground object categories in the process of smoothing ground object categories,resulting in the mixing of ground information.The Discontinuity Preserving Relaxation(DPR)algorithm is able to smooth hyperspectral remote sensing images at the same time,Taking into account the problem that the boundary categories of ground features tend to be the same.Based on the above several algorithms,this paper proposes a new dimension-reduction scheme for feature selection of hyperspectral remote sensing images.In this scheme,in the first step,the FCM subspace division method is used to divide the hyperspectral data set into wave segment subspace.Then,Maximum Entropy(ME)is taken as the evaluation criterion.The band selection is carried out on the data set divided by wave segment subspace and a band subset containing the most abundant feature information is obtained.Then,the band subset is smoothed and post-processed by DPR.The Support Vector Machine(SVM)was used to classify the smoothed band subsets.In this paper,three internationally common hyperspectral remote sensing image data sets,namely Indian Pines data set,Pavia University data set and Salinas data set,were used to verify the proposed scheme,and the experimental results proved the effectiveness of the scheme.