Radar and optical data comparison/integration for urban delineation: A case study

This study compared spaceborne radar and radar-derived products with optical data for urban delineation. RADARSAT and Landsat Thematic Mapper (TM) multispectral data were assessed independently and in combination. The primary methodology was supervised spectral signature extraction and the application of a maximum-likelihood statistical decision rule to classify surface features in the Kathmandu Valley, Nepal. Relative accuracy of the resultant classifications was established by comparison to ground-truth information. Both radar post-classification smoothing and Variance texture measures were improvements over the poor results achieved with the unfiltered, original radar data. Speckle reduction procedures were found to be very advantageous. Combinations of radar-derived products greatly improved results, achieving an overall accuracy of nearly 90 percent. The best overall accuracy was achieved with the merger that included a texture image derived from despeckled radar and the despeckled original radar. The radar and radar-derived combination achieved much better results than did the TM and were comparable to a combined radar and TM data classification. The systematic strategy of this study, determination of the best individual method before introducing the next procedure, was effective in managing a very complex, almost infinite set of analysis possibilities.