Potential of Semi-automatic Object-based Land Cover Classifications using Very High Resolution Satellite Images: Tetuan-city Comparison Case Study

Abstract

Nowadays, mapping agencies need to update their topographic and information systems databases in order to draw: (i) urban planning (e.g., urban sprawl, road networks, etc.), and (ii) environmental planning (e.g., ecological corridors, forest change, etc.). In this context, object-based land cover classifications using very-high resolution satellites (VHRS) images (e.g., IKONOS, QuickBird, GeoEye WorldView-2, and recently Pléaides) are interesting tools to satisfy the needs of both scientific community and decision makers in local authorities. Accordingly, the number of non-experts in using VHRS remotely sensed images is increasing, and the requirements of semi-automated/automated processing using less number of parameters is an expanding need. Recently, object-based classification algorithms have been integrated to commercial remote-sensing software. These last, are trying developing user-friendly automated classification solutions. In this research, an evaluation and comparison study of the implementation of these object-based land-cover classifications using VHRS – integrated in eCognition Developer, ERDAS, and ENVI – has been carried out. Potential of these classification methods in the case of Tetuan-city (Northern Morocco) has been carried out by evaluating: (i) the number of required parameters as imperative inputs, and (ii) the accuracy assessment of land cover classifications. In this case, eCognition Developer shows promising results with: (i) two input parameters; and (ii) Kappa coefficient mean value of about 0.80.

Keywords: Land cover, object-based classification, Tetuan, VHRS