Comparing feature extraction techniques for urban land-use classification

Extraction of a reliable feature and improvement of the classification accuracy have been among the main tasks in digital image processing. Over the years, many techniques have been developed and tested for processing and analysis of multi-spectral image data with fewer dimensionalities. Although it is desirable, the error increment due to the reduction in dimensionality must be constrained to be adequately small. Finding the minimum number of feature vectors, which represent observations with reduced dimensionality without sacrificing the discriminating power of pattern classes, along with finding the specific feature vectors, has been one of the most important problems in the field of pattern analysis. In this study, the conventional statistical principal component analysis and self-organizing feature map of artificial neural network techniques were used in order to reduce the volume and to maximize information content of input data. The results were compared for their effectiveness in land-use analysis.