earticle

영어

This work proposes the application of a hybrid Particle Swarm Optimization (PSO) with Levenberg Marquardt Back-Propagation (LMBP) algorithm to train Artificial Neural Networks (ANNs) for classification of medium resolution multispectral satellite imageries. ANNs have been widely used in satellite image classification and have been shown to outperform traditional classifiers in many situations. However the Back Propagation (BP) algorithm traditionally used in training ANN suffer the problem of local minima entrapment, thus affecting the accuracy and performance of the ANN classifier. A hybrid combination of PSO and LMBP algorithm is applied to resolve the aforementioned problem and enhance the accuracy and performance of the ANN classifier. To investigate the performance of the proposed method, medium resolution multispectral satellite imagery was classified using the proposed classifier and its performance compared with that of conventional LMBP and Scaled Conjugate Back-Propagation (SCBP) trained ANN classifier. Results obtained shows that the hybrid PSO and LMBP trained ANN classifier out performs the conventional LMBP and SCBP trained ANN classifier and achieves ≈95% accuracy on the test medium resolution satellite imagery.