摘要:Speckle is a characteristic random noise from coherent imaging systems like synthetic aperture radar (SAR). Due to this noise, even areas with sharp features in continuous surfaces (e.g., crevasses on glaciers) will be characterized in SAR images by grainy texture and high variation in digital numbers. Therefore, the speckle must be reduced before SAR images can be used for measuring glacier velocities by cross-correlation algorithms. To solve this problem, four approaches based on adaptive filtering (i.e., Lee filter) were tested for data pre-processing prior to extracting the velocity fields from glaciers. The Lee filter was used in two ways: (i) one-pass and (ii) two-pass filtering. Furthermore, two parameters were used with the Lee filter to explain the data variability: number of looks and standard deviation of the scene. Results evaluation was carried out comparing the velocity vectors resulting from original and filtered images with published data on dynamics of the glaciers in the northern part of the Antarctic Peninsula. In terms of speckle suppression, all approaches yielded positive results. However, the two-pass filter does not preserve the crevasses edges and the resulting images are not considered for the final result comparison. In general, images processed with one-pass filter showed better results for extraction of velocity vectors with the cross-correlation algorithm than the original ones, and were accepted for an automatic processing chain to derive dynamic parameters of glaciers. Furthermore, resulting velocity vectors agree with published data and show a slight increase in velocity between 2001 and 2005.
关键词:Keywords: Glacier Dynamics, SAR Remote Sensing, Feature Tracking.