The Classification Of Remote Sensing Image Based On GA-PSO Optimize Support Vector Machine

Remote sensing image classification is one of the key technologies in application field of remote sensing technology.High precision remote sensing image classification algorithm is the premise to solve practical problems.Domestic and foreign scholars have applied numerous machine learning algorithms to the classification of remote sensing images,in order to improve the classification accuracy of remote sensing images and to excavate more effective information.Support vector machine(SVM)algorithm is a popular classification algorithm in recent years.It is proposed on the basis of the VC dimension theory of statistical learning theory and the principle of minimum structure risk.Compared with other machine learning algorithms,the algorithm has a unique advantage,especially in dealing with small samples and nonlinear problems,the classification results are very significant,especially.Therefore,SVM can maintain a high classification accuracy in the relevant theoretical research experiments and practical applications.At present,the algorithm is widely applied in various fields.In this paper,based on the SVM classification algorithm,combined with the more widely used intelligent optimization algorithms,the parameters of SVM are optimized to improve the classification accuracy of remote sensing images.Based on the in-depth study of the theory of particle swarm optimization(PSO)and genetic algorithm(GA),this paper analyzes their advantages and disadvantages respectively.It is found that when the particle swarm optimization(PSO)is used to optimize the parameters of the SVM,there are problems such as premature convergence,low classification accuracy and easy to relapse into local best.In this paper,a hybrid optimization algorithm(GA-PSO)is proposed.It is based on the adaptive weighted particle swarm optimization(PSO)and a genetic algorithm cross mutation operator is introduced,The improved algorithm is used to optimize the parameters of SVM classifier and classify remote sensing images.The paper takes the high-resolution remote sensing image Quick Bird and the medium resolution remote sensing image Landsat8 as examples,do the preprocessing of image cutting,image fusion,etc.Then we use the PSO-SVM algorithm and the new algorithm built in this paper(GA-PSO-SVM)to conduct land use classification experiments,at the same time,the classification accuracy of the two algorithms is compared and analyzed.The results show that the average classification accuracy of Quick Bird remote sensing images is increased by 5.07%,and the average classification accuracy of Landsat8 remote sensing images is increased by 6.32%.Experiments show that the new algorithm to improve the search performance of particle swarm optimization.It is an effective way to find the best SVM classifier parameters and get high classification accuracy.