Reduction of layover distortion in SAR imagery

Abstract

Layover distortion in synthetic aperture radar (SAR) imagery is interpreted as the result of the inability of an obliquely oriented remote sensor to distinguish the horizontal and vertical components of its range measurement. A geometric model of the SAR imaging process is employed in conjunction with digital terrain elevation data to simulate the layover distortion inherent in a Seasat SAR image of a mountainous scene. The simulated layover database is then used to solve the inverse imaging problem, which is to reduce layover distortion by geographically redistributing concentrations of the SAR image amplitude that can be attributed to the focusing influence of mountain slopes that face the radar. Cross-correlation and classification statistics are used to compare the results with an aerial photograph. Two classes of terrain are distinguished. These correspond to positive and negative correlations of radar and optical reflectivities for different surface textures. They also correspond to higher and lower scene elevations. These two classes of terrain reflectivity are misregistered by about 160 m along the SAR track and are separated by about 60 m in mean elevation. The misregistration is shown to be consistent with the clutterlock and autofocus procedures employed by the SAR scene correlator as it responds to the variance in the earth's relative rotational velocity over the three-dimensional terrain.