Textural analysis for urban class discrimination using IKONOS imagery
Other titre : Analyse texturale pour la discrimination des classes urbaines sur des images IKONOS
High spatial resolution imagery can be a very significant source of detailed land cover and land use data necessary for better urban planning and management, which is becoming increasingly important due to the growing human population. However, traditional methods, based on spectral data, used to extract this information from remote sensing imagery have proven to be unsuitable for high-resolution images. Spatial data, or texture, has been widely investigated as a supplement to spectral data for the analysis of complex urban scenes. However, the application of these techniques on high spatial resolution imagery, such as those obtained by the IKONOS satellites, has yet to be studied. This research, therefore, focuses on the extraction of texture features through the use of the Grey Level Co-occurrence Matrix texture analysis technique, which are then combined with the spectral data in the Maximum Likelihood Classification approach, as a method for obtaining more accurate urban land cover and land use information from high spatial resolution IKONOS imagery. In this study, classifications were done using three datasets: a spatial dataset consisting of three texture channels (Mean, Homogeneity and Dissimilarity), a spectral dataset consisting of four spectral channels (Red, Green, Blue and N-IR), and a combination dataset (spatial and spectral). The results show that the spatial dataset produced an overall classification accuracy of 73.5%. The spectral dataset produced a slightly higher overall classification accuracy of 78.9%, an increase over the spatial dataset of 5.4%. The combination dataset produced the highest overall classification accuracy of 86.1%, which is an increase of 7.2% over the spectral dataset. These results demonstrate great potential for the contribution of texture and high-resolution images in deriving more accurate and detailed urban information.