A transferable and stable object oriented classification approach in various urban areas and various high resolution sensors

Abstract

High resolution imagery enables a detailed analysis of the small scale and heterogeneous urban environment. A modular object-oriented classification methodology has been developed for the sensors IKONOS and QuickBird. The paper focuses on a modular concept that aims at stable transferability with fast and easy adjustment possibilities on the particular urban structure or the particular spectral features. The chronologic workflow from the segmentation optimization, the fuzzy concept of classification and the final classification results are presented. Study areas are the opposed urban areas of megacity Istanbul, Turkey and the planned small town of Wuda, China. A validation has been carried out for both urban areas for an assessment of the methodology and a performance of the transfer.