A Texture Measure Defined Over Intuitionistic Fuzzy Set Theory for the Detection of Built-Up Areas in High-Resolution SAR Images

Visual interpretation of built-up areas in high-resolution (HR) SAR images usually relies on the strong backscattering behavior and the heterogeneity. The bright pixels characterizing the built-up areas can be easily identified, whereas the built-up pixels appear in medium and low intensity may be missed. The coefficient of variation traditionally provides a way to discriminate heterogeneous and homogeneous regions. The analysis of the statistics of the coefficient of variation on different land cover classes allows us to construct two membership functions (MFs) representing the built-up class and non-built-up class, respectively. Then, each pixel is attached with a membership degree to either of the two classes computed by using the established MFs. At this level, the amplitude of the SAR image is transformed into the intuitionistic fuzzy sets (IFSs). The context texture information is specified by the labeled similarity matrix (LSM) computed using the IFSs operators. Since different IFSs operators are used, the proposed technique can be implemented in different ways. The proposed techniques were tested on a collection of 11 images selected from TerraSAR-X images acquired on Nanjing (China) and a COSMO-SkyMed image acquired on Hangzhou (China). By comparison with the existing techniques, the effectiveness of the proposed techniques in identifying and detecting built-up areas was confirmed.