Extract Profile of Urbanized Area in Beijing Assisted by Skewness

In Beijing, some crop field is similar to urbanized area in spectral space of ETM+ image, and this is an obstacle to extract urbanized area of Beijing from ETM+ image. Furthermore, classical texture such as variance can not improve the seperability of such two landcovers. This article proposed a skewness based framework to extract the profile of urbanized area, while skewness was seldom applied in optical images. Firstly, skewness image was generated from the panchromatic image of ETM+, and it was smoothened using low-pass filter. Secondly, skewness and spectral information, from multi-spectral bands, were combined to fully utilized both texture and spectral information. Finally, morphological operation and DEM data were employed to perfect the classification result. Through visual interpretation, the extracted urbanized area is accurate in large scale, so the “profile” of urban was obtained. At last, accuracy assessment showed that accuracy of the skewness based classification was 10% higher than traditional approaches.