Research On Hyperspectral Image Unmixing Method Based On Salp Swarm Algorithm And Shadow Model

Remote sensing images can provide information of natural resources and environment.However,hyperspectral image unmixing is an important technique to analyze the feature information of remote sensing images.Because features in the natural environment are complex and hyperspectral images are easily affected by the atmosphere,the analysis of hyperspectral images becomes complicated.In order to identify features more accurately,hyperspectral image unmixing technique is becoming particularly important.The thesis proposes a hyperspectral image unmixing method based on salp swarm algorithm and shadow multilinear mixing model,aiming to achieve higher unmixing accuracy to estimate the proportions of features accurately.First of all,salp swarm algorithm is one of the most recently proposed intelligence optimization algorithms.Compared with other intelligence optimization algorithms,the optimization strategy of salp swarm algorithm needs to be improved to enhance convergence accuracy and convergence speed.The thesis proposes a salp swarm algorithm based on reduction factor and dynamic learning.Simulation results show that the proposed algorithm has great performance on convergence accuracy and convergence speed.Secondly,shadow multilinear mixing model is a effective model for hyperspectral unmixing problem,which contains general properties of multi-spectral data and shadow effect.In this thesis,shadow multilinear mixing model is used to reconstruct hyperspectral data,so that hyperspectral unmixing problems are transformed into optimization problems,avoiding complicated calculation of traditional hyperspectral unmixing methods.Finally,the improved salp swarm algorithm is used to solve the objective function to achieve higher unmixing accuracy.At the same time,the two constraints of abundance can be satisfied by controlling the search boundary of the improved salp swarm algorithm,which could simplify the steps of unmixing methods.The results of simulation test show that the proposed hyperspectral unmixing method has good anti-noise capability and has robust performance on different pixel purity data.The results of real images test show that the proposed hyperspectral unmixing method can improve unmixing accuracy.