Texture analysis of urban areas in ERS SAR imagery for map updating

In single-band and single-polarized SAR image classification, textural information is important, both for pixel and segment based classification schemes. To study the map updating capabilities of such sensors in urban areas, several texture measures were studied. Among them are statistical measures, wavelet energy, fractal dimension, lacunarity, and semivariogram. The latter was chosen as an alternative for the well known gray-level co-occurrence family of features. Two urban areas were studied using ERS1/2 data, one of which is reported: the conurbation around Rotterdam and The Hague in The Netherlands. The area can be characterized as a well-planned dispersed urban area with residential areas, industry, greenhouses, pasture, arable land, and some forest. The digital map is a 1:250,000 vector map (VMapl). The texture measures that gave the best landcover separability for this area are: mean intensity, variance, skew, weighted-rank fill ratio, semivariograms (or alternatively wavelet energy measures), and lacunarity. The latter is preferred to be included in case more than one man-made landcover class is involved, which is often the case in urban environments