Urban environments are very dynamic phenomena, and it is essential to update urban-related information for various applications. In this regard, remotely sensed data have been utilized widely to extract and monitor urban land use and land cover changes. Particularly, synthetic aperture radar (SAR) data, due to several advantages of this technology in comparison to passive sensors, provides better performance especially in tropical regions. However, the methodological approaches for extraction of information from SAR images are another important task that needs to be considered appropriately. This paper attempts to investigate and compare the performance of different image classification techniques for extracting urban areas using advanced land observing satellite phased array type L-band synthetic aperture radar imagery. Several object- [such as rule based (RB), support vector machine (SVM) and K-nearest neighbor (K-NN)] and pixel-based [decision tree (DT)] classification techniques were implemented, and their results were compared in detail. The overall results indicated RB classification performed better than other techniques. Furthermore, DT method, due to its predefined rules, distinguished the land cover classes better than SVM and K-NN, which were based on training datasets. Nevertheless, this study confirms the potential of SAR data and object-based classification techniques in urban detection and land cover mapping.