Impervious surface area extraction from IKONOS imagery using an object-based fuzzy method

The study of impervious surfaces is crucial to the sustainable development of urban areas due to its strong impact on urban environments. Remotely sensed high-resolution imagery has the advantage of providing more spatial details; however, digital image processing algorithms have not been well developed to accommodate this advantage and other characteristics of such imagery. In this article, an object-based fuzzy classification approach for impervious surface extraction was developed and applied to two pan-sharpened multi-spectral IKONOS images covering the residential and central business district (CBD) areas of Indianapolis, Indiana, USA. Fuzzy rules based on spectral, spatial and texture attributes, were developed to extract impervious surfaces. An accuracy assessment was performed for the final maps. The results indicated that the spatial patterns of extracted features were in accordance with those in the original images and the boundaries of features were appropriately delineated. Impervious surfaces were extracted with an accuracy of 95% in the residential area and 92% in the CBD area. Road extraction achieved accuracy a bit lower, with 93% and 90% from the residential and CBD area, respectively. Buildings were extracted with an accuracy of 94% from the residential area while 89% from the CBD area. It is suggested that the CBD area had a higher spectral complexity, building displacement and the shadow problem, leading to a more difficult estimation and mapping of impervious surfaces.