High Resolution Remote Sensing Image Classification Based on Improved Particle Swarm Optimization Optimized BP Neural Network

Abstract

The classification efficiency and accuracy of high resolution remote sensing image classification method based on BP neural network or Particle Swarm optimization (PSO) optimized BP (PSO-BP) neural network are affected because the weight threshold parameter convergence of BP neural network is slow, and the inertia weight and learning factor of standard PSO are fixed. In order to solve these problems, a classification method based on Improved Particle Swarm optimization (IPSO) optimized BP (IPSO-BP) neural network for high resolution remote sensing image is proposed, which the inertia weight and learning factor of PSO can change linearly and dynamically with the increase of iteration times. The classification model is constructed by using the digital matrix and color feature vector of remote sensing image as input and output training. The optimal weight threshold obtained by iterative optimization of IPSO is directly assigned to BP neural network, and then the remote sensing image classification is realized. The experimental results show that, compared with the maximum likelihood, support vector machine, object-oriented nearest neighbor, BP neural network, and PSO-BP neural network remote sensing image classification methods, the proposed method has significantly improved the classification accuracy of single category, overall classification accuracy and kappa coefficient, reflecting the good classification effect.