Abstract
In this paper a practical method is demonstrated for estimating topography of natural terrain from the radiometric, or shading, information in a synthetic aperture radar (SAR) image. While this problem has been considered before for radar, methods available in the computer vision literature have not previously been utilized in its solution. We treat this as a computer vision problem, viz. shape from shading (SFS). A review of the relevant characteristics of SAR imagery is presented followed by a formulation of the SFS problem for SAR. Because of the noise inherent in SAR imagery a cost minimization approach is used which allows for noise and incorporates a regularization term in the cost function. Previously developed numerical methods are adapted to SAR imagery by incorporating radar reflectance models and by solving for surface slopes in a rotated system of coordinates—one which represents surface height relative to a plane parallel to the line-of-sight. Two difficulties are recognized. First, unknown reflectance model parameters must be estimated from the image data. Second, shading provides reliable information about the high frequency components of the surface but not the low frequency components. Both difficulties are reduced if auxiliary low resolution surface height information is available. This is demonstrated by combining Shuttle Imaging Radar-B (SIR-B) SAR images with much lower resolution terrain elevation data to construct high resolution terrain elevation estimates. The estimation of Venusian surface topography from Magellan SAR imagery is discussed and methods are suggested for combining shading information with geometric stereo.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call