Feature extraction and classification of urban high-resolution satellite imagery based on morphological preprocessing

Abstract

Classification of panchromatic high resolution data from urban areas using a three-step approach based on morphological preprocessing is investigated. First, the morphological composition of geodesic opening and closing operations of different sizes is used in order to build a morphological profile. Secondly, feature extraction is applied in the second step. Thirdly, statistical classifiers are used to classify the features. Examples of the application of the proposed method are given for one satellite high-resolution data set from Athens, Greece. Both discriminant analysis (DA) and decision boundary feature extraction (DBFE) are applied successfully in the feature extraction phase. For the statistical classification, original, leave-one out (LOO), and enhanced statistics are used and evaluated. In experiments, the use of DA and DBFE shows promise when used with original and LOO statistics.