Conjugate gradient method neural network for medium resolution remote sensing image classification

Neural network (NN) classification has found wide use in remote sensing applications. There are a large number of ANN types available, and each focus on improving different classification performance. The conjugate gradient method is one of the efficient and low memory requirement methods. In this paper, a neural network using conjugate gradient method classifier is employed to classify three components derived by using principal component analysis to original six bands Landsat TM images. Comparison with a conventional classifier shows this NN performs better in both visualization inspection and quantitative evaluation.