Water Extraction of Airborne Polarimetric SAR by Introducing Eigenvalue Relative Difference

In high-resolution airborne SAR images, water bodies, paddy fields, roads, bare soils and other ground objects exhibit similar weak scattering characteristics. At present, the water body extraction algorithm based on the weak echo characteristics cannot eliminate the interference of other ground objects, resulting in unsatisfactory water extraction results. Aiming at the KU-band airborne polarimetric SAR image, this paper proposes a fine extraction method of airborne polarimetric SAR water with the relative difference of single-reflection eigenvalues. Firstly, the H-α Wishart classification algorithm is used to classify the weak scattering features, and then, the weakly scattered features are re-segmented based on the pre-classification results to construct the object unit. Finally, based on the single-reflex eigenvalue relative difference feature, the threshold method is used to realize the water fine extraction. Experiments show that the relative difference of single-reflection eigenvalues can effectively separate the water body from the rest of the weakly scattered features, and the object unit constructed by the fractal network evolution algorithm can obtain better water body fineness than the simple pixel-level analysis.