This thesis describes the study of Artificial Neural Network (ANN) based techniques for the classification of aerial images for various types of land-use. In this study both gray-scale and multispectral aerial images were used in land-use classification. Three approaches were used for the preparation of the data as inputs to the ANN, including histograms of the pixel intensities, textural parameters extracted from the image, and matrices of pixels for spatial information. The approach using textural parameters was found to be the best for both gray-scale and multispectral image classification. A probabilistic neural network was employed. A high level of accuracy was achieved with both gray-scale (92%) and multispectral images (89%).
