Vol. 9 No 6 December 2001
Synthetic aperture radar (SAR) being an active sensor can image during day and night, and has all-weather capability. SAR images are characterised by shadows, highlights, corner reflector effect, layover and shortening. Being sensitive to surface roughness and topography, it can easily detect constructed areas/structures and camouflaged objects. It has advantage of low revisit time due to steerable beam and cloud penetration capability. SAR being a coherent imaging system preserves phase information. The phase information of a pair of single look complex (SLC) image is used for the generation of digital elevation map (DEM).
SAR interferometry is a spin-off of active microwave remote sensing, in particular SAR technique. In this technique, the phases of two pixels gathered from two complex SAR images of the same area are combined to give information of terrain elevation. This technique is particularly useful for generating digital topographic map (DTM) of those areas that are constantly under cloud cover. A semi-automatic software capable of generating accurate DTM from a pair of SLC image has been developed by the DRDO. First, the SLC image pair is registered to the order of one tenth of pixel accuracy and then two registered SLC images are used to generate the interferogram. The interferogram is then filtered and flattened after baseline modelling. Phase unwrapping is done to remove the modulo 2¶ effect. The unwrapped interferogram is then converted to DTM by height modelling.
Speckle Reduction Filters
SAR images suffer from speckle noise, which
is nothing but random interferometric fringes and is multiplicative in nature. This
speckle noise has to be removed before processing and interpretation of SAR images. A
software has been developed for mean filter, sigma mean filter, median filter, pivotal
median filter, rational filter, Lee filter, Kuan filter, and frost filter for the removal
of speckle noise. Measures, such as random noise reduction index and edge retention index
have been developed for
SAR images are very rich in textural information and are suitable for the study of snow/ice properties. A software has been developed which utilises the water equivalence and backscattering property along with some texture-based measures, like contrast, variance, and homogeneity for the segmentation of glaciers in various classes of snow.
Urban Area Extraction
SAR images have shown a significant potential in detection of man-made targets. The images have inherent properties, like corner reflector effect and local distribution of gray levels, which are very useful for discrimination of man-made targets. A software has been developed for the detection of urban area using two parameters, viz., Weibull parameter of local window histogram and windowed fractal dimension. The two measures are then used in synergy to extract urban area.
SAR Image Simulation
DRDO has developed a software for the simulation of SAR images in different imaging geometry (strip mode, circular scanning mode and spotlite mode), different pixel spacing, different sensor parameters (height, frequency, etc.). It uses the IRS 1C/1D LISS/PAN images, DEM and ground truth as input and can simulate SAR images. It first calculates the backscattering coefficient map using complex algorithms and then redistributes it as per the sensor acquisition geometry.