A Closed-Form Expression Relating Classification Accuracy to SAR System Calibration Uncertainty

摘要

The choice of synthetic aperture radar (SAR) system design parameters such as radiometric calibration uncertainty and noise-equivalent sigma zero has a significant impact on applications exploiting SAR image data. However, methods for quantitatively translating the impact of system and mission parameter choices into the application context are lacking. This letter addresses this subject and derives a closed-form algebraic expression for translating radiometric biases due to calibration uncertainties—an important SAR engineering specification—into estimates of classification uncertainties for the restricted case of two homogeneous classes characterized by different backscatter intensity levels. The expression is exploited within this letter to estimate the potential impact of radiometric uncertainty on SAR-derived thematic maps. The results indicate that classification results based on the joint use of data from future SAR constellations in particular may be significantly degraded due to calibration uncertainties.