THE COMPARISON OF CLASSIFICATION METHODS WITH ANN AND MAXIMUM LIKELIHOOD

Artificial Neural network (ANN) models are a powerful and reasonable alternative to conventional classifiers. In this paper, we compared ANN classifiers and conventional classifiers (Maximum Likelihood). The spectral data used in the project is from the 12-band airborne multi spectral scanner system and the target image covers southern part of TippecanoeCounty in Indiana, US. First, conventional classification was accomplished by using the software named “MultiSpec”. Second, artificial neural network was designed to classify land use by using so called ‘back propagation method’. In first and second phase, we used same training data and test data. Third, both two results are compared about the accuracy. Based on trained data, we also classify on the whole area. By doing this we proved the usefulness of neural network classification system about the high dimensional data. The experimental results shows that ANN and Maximum Likelihood methods produce similar overall accuracy. However, applying ANN takes a lot of time comparing with Maximum Likelihoo