FOREST TYPE CLASSIFICATION USING MORPHOLOGICAL OPERATORS AND FOREST PA METHOD

Abstract

Morphological operators have obtained great attention in the fields of image processing and pattern recognition due to the providing relevant spatial information for the classification. This study evaluates the impacts of morphological operators from dual-polarized, Advanced Land-Observing Satellite (ALOS) and Phased Array L-band Synthetic Aperture Radar (PALSAR) data for the classification of forested areas. To this aim, the opening and closing operators with different size of structuring elements were exploited as morphological features in our study. For the classification of the forested areas, three different classification models (Support Vector Machines, Random Forests and Forest PA) were performed. The experimental results show that morphological features have increased the classification accuracy by 6.82% and 10.36% for Forest PA (Forest by Penalizing Attributes) and Random Forests (RF), respectively, while decreased the accuracy by 1.43% for Support Vector Machines (SVM). Furthermore, Forest PA yields the highest accuracy with morphological features followed by RF.