Comparative Study of Object Based Image Analysis on High Resolution Satellite Images for Urban Development

For land cover classification and urban area analysis remote sensing techniques are gaining more importance. Numerous remote sensing techniques have been developed for analyzing the satellite images. The launch of Worldview-2 satellite with resolution of 0.5 m signaled the advent of high resolution satellite images. Such images offer an exciting possibility for feature extraction as well as complex land cover classification. The most suitable approach for analysis of high resolution satellite images is Object based Image Analysis (OBIA). OBIA is relatively new class of algorithm that have been developed to focus not only on the spectral properties of features but also spatial such as their shape, orientation, texture, contextual relation feature and so on. In this study, authors have used WV-2 image and classification is carried out with two approaches. First method is based on rule set approach where domain expert knowledge is represented in rules with Cognition network language whereas second approach is Nearest Neighbor (NN) classification. Accuracy of classification carried out with the help of confusion matrix which indicates that rule based classification is more accurate as compare to NN. However for complex land like urban area both approaches are suitable as compare to pixel based approach.