Sentinel-1 Spatially Varying Maximum-Likelihood Coherent Change Detection

Traditional sample complex correlation change detection (CCD) for synthetic aperture radar decorrelates and thus exhibits false change in areas of low clutter-to-noise power ratio. The complex reflectance change detection (CRCD) maximum likelihood estimator uses a noise power estimate for each SAR image to lower false change. This paper extends the CRCD metric to use spatially varying noise power estimates. The Sentinel-1 SAR constellation is a publicly available data source operating in an interferometric mode with a wide 250km swath width. The noise power varies significantly across each Sentinel-1 image. The CRCD estimate with spatially varying noise (SV-CRCD) is applied to Sentinel-1 flood mapping. Accuracy improvements are demonstrated over CCD due to SV-CRCD rejecting false change in regions of low return.