An Integrated Image Matching and Block Adjustment Method for Regional-Scale Spaceborne SAR Image Mapping

Abstract

The rapid development of spaceborne synthetic aperture radar (SAR) systems has provided abundant data for Earth observation. However, their initial relative geometric positioning accuracy remains insufficient for regional-scale SAR image mapping. To address this challenge, an integrated method combining image matching and block adjustment is proposed. First, regional-scale SAR image matching is decomposed into parallelizable slice image matching based on grid division of geographic overlapping areas. An enhanced SAR-scale-invariant feature transform (SIFT) (ESAR-SIFT) algorithm is developed to extract reliable tie points (TPs) from slice images while preserving more edge details and mitigating neighboring feature interference. Next, the extracted TPs are mapped to the whole SAR image through a geometric transformation model, and a joint geometric–radiometric constraint-based adaptive nonmaximum suppression algorithm is proposed to achieve a spatially homogenized distribution of TPs. Finally, using these TPs, a digital elevation model-aided planar block adjustment model is established to estimate the orientation parameters of all images and generate orthorectified images. Experiments on 16 LuTan-1 (LT-1) and 4 GaoFen-3 (GF-3) images validate the effectiveness of the proposed method. The ESAR-SIFT algorithm demonstrates superior performance in both matching quantity and accuracy across diverse terrains. The root mean square errors of homogenized TPs are reduced from 4.034 pixels to 0.731 pixels for LT-1 and from 3.567 pixels to 0.702 pixels for GF-3. The resulting relative positioning accuracies reach 0.56 m for LT-1 (3-m resolution) and 8.43 m for GF-3 (10-m resolution). These results confirm that the proposed method enables accurate and efficient regional-scale SAR image mapping.