Abstract
Comparison of an Artificial Neural Network (ANN) classifier with Maximum Likelihood (ML) classifier for different land cover classification has been examined previously using different sensors such as Landsat TM, SPOT, AVHRR and SAR. In this study, the Indian Remote Sensing (IRS‐1C) satellite sensor LISS‐III (Linear Image Self‐Scanner) was used to compare these classification methods for forested land. Training samples for classification and accuracy assessment were collected through extensive field surveys using a false colour composite image. The same training areas were used for both classifiers to classify the entire forest land of Banavasi Range of the Western Ghats, India using EASI/PACE (PCI) image processing software. A detailed accuracy assessment was done for both classifiers using existing forest survey maps and information collected from the field survey. Results showed that the artificial neural network has a slightly better performance in discriminating forest plantations and water, but no significant difference in classifying the homogenous natural forest, when compared to the maximum likelihood classifier.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call