Study of Image Classification accuracy of Supervised Neural Network with Traditional Statistical Methods for Multi-temporal Remotely Sensed Data

摘要

Traditionally statistical procedures have been developed such as supervised, unsupervised and hybrid schemes for classifying the multi- spectral image data. ANN is a computational tool for classification as well as clustering. Multi-temporal remote sensing data are widely used instead of single date imagery for the purpose of vegetation survey, since advantage can be taken of variation in the phenology of different vegetation type. It is only possible, however, to monitor the dynamic behavior of vegetation through the seasons by having remote sensing data available on a regular basis. This paper presents comparative study of accuracy assessment in classification of multi temporal RS data of Hardoi (UP) using two approaches viz conventional statistical classifier-maximum likelihood classification (MXL) and neural network.