Analysis of coherence images over urban areas in the extraction of buildings heights

In this paper we show how ERS coherence imagery can be useful to extract information about building heights in urban areas. The first part analyses the causes of decorrelation over urban areas. Spatial decorrelation is partially offset by spectral shift filtering, while thermal and processor decorrelation can be neglected. In order to extract building heights, we use tandem coherence only, so that temporal coherence can be set equal to one. Hence, the estimated coherence over built-up areas coincides with volume scattering decorrelation. Using the Van Cittert Zernike theorem, we introduce a model for the backscatter coefficient of urban areas. This model links volume decorrelation to a height variable so that theoretical coherence values can be obtained for varying heights. At this stage, it is possible to compare theoretical and experimental values of tandem coherence and mean heights of buildings can be extracted. To test the procedure, four sites in Italy, have been considered. For one area, results show that in situ measurements are in agreement with values of building heights obtained from coherence comparison. For two areas, results seem to agree with local observations. For the fourth area, discrepancies occur, probably due to weather effects. Hence, the described procedure seems to be promising and more accurate results may be achieved using fine resolution coherence images. However, the effects of weather on tandem coherence in urban areas should be investigated and taken into account.