The comparison of Artificial Neural Network to and maximum likelihood algorithms for forest changes detection

Remote Sensing Technology is considered one of the most important sources of spatial and thematic data in the developed world of today. The objective of this work is a comparison of two different methods of change detection in forests using Landsat images. Therefore, sensor Landsat TM images of 1990 and 2011 (ETM+) satellite images have been used.

In the classification of images, the maximum likelihood algorithm, and artificial neural network to multilayer perceptron method were used.

Evaluated results showed that the algorithm approach, the maximum likelihood overall accuracy, and kappa coefficient maps classified in TM image, respectively, are 96.72 and 0.96 percent and image ETM+ 98.02 and 0.97 percent, and the method of artificial neural networks, overall accuracy and kappa coefficient map classified, TM image was 98.22 and 0.97% and ETM+ image was 98.34and 0.97 percent respectively. Following TM and ETM+ classification maps to detect the changes were marked and the map changes obtained.

The results of this study showed that using Landsat data along with data from have inventory capabilities of forest change mapping