This paper assesses the utility of a variety of polarimetric interferometric SAR data sources for the identification of ground pixels over a wooded area to enable accurate digital terrain model generation from the InSAR height of the selected ground hit pixels. The information sources assessed include the radar backscatter, interferometric coherence, surface scattering proportion (based on Freeman-Durden decomposition), InSAR standard deviation and a model of expected height error; all of these data sets are integral parts of a fully polarimetric SAR interferometry data set, but are poorly utilised currently, and often ignored once the initial DTM has been produced. The method is applied to Monks Wood, a small semi-natural deciduous woodland in Cambridgeshire, using airborne E-SAR data collected in June 2000. The results are validated against theodolite data and a LiDAR-derived DTM. The results show that increasing the amount of data going into the DTM creation does not necessarily increase the accuracy of the final DTM. Instead, the most accurate method, for the whole wood, was the simplest method, namely the application of a fixed window minimum filtering algorithm, followed by a mean filter. The results suggest that the signal-to-noise error is too high within the InSAR data set for the more sophisticated solutions to work, due to the propagation and accumulation of errors. However, for a subset of the area using ancillary data to identify ground pixels out-performs the minimum filtering method.
展开▼
