Optimal classification of remote sensing data using directed and non-directed instruction of neural networks

Abstract<br /> The study of multi classification of data has become one of the important issues which geographical studies focus on especially those which take their data from satellites.<br /> Two ways of classification has been taken into consideration, each one completes the other. The two ways are used together to get the benefits of both and to obtain the full advantage. The data of remote sensing were chosen to be tested. <br /> The system suggested is a software package which consists of neural networks of supervised learning and neural networks of unsupervised learning to classify the land of (Al-Mosul Dam) in Mosul city.<br /> The best use of remotely sensed data is by using the methods of supervised and unsupervised classification consequently which improve the primary input data in the classification. Thus, a high degree of accuracy and efficiency in the classification are obtained. <br /> It is worth mentioning that software package used in the integration consists of the nets: the supervised ART-II neural network and the unsupervised Kohonen neural network.