Glacier Surface Motion Estimation from SAR Intensity Images Based on Subpixel Gradient Correlation

Abstract

With the current extensive availability of synthetic-aperture radar (SAR) datasets with high temporal (e.g., a repeat cycle of a few or a dozen days) and spatial resolution (e.g., in the order of ∼1 m), radar remote sensing possesses an increasing potential for the monitoring of glacier surface motion thanks to the nearly weather and time-independent advantages. This paper proposes a robust subpixel frequency-based image correlation method for dense matching and integrates the improved matching into a workflow of glacier surface motion estimation using SAR intensity images with specific pre-processing and post-processing steps. The proposed matching method combines complex edge maps and local upsampling in the frequency domain for subpixel intensity tracking, which ensure the accuracy and robustness of glacier surface motion estimation. Experiments were carried out with TerraSAR-X and Sentinel-1 images covering two glacier areas in pole and alpine regions. The results of the monitoring and investigation of glacier motion validate the feasibility and reliability of the presented motion estimation method based on subpixel gradient correlation. The comparative results using both simulated and real SAR data indicate that the proposed matching method outperforms commonly used correlation-based matching methods in terms of matching accuracy and the ability to obtain correct matches.

Long-Term Monitoring and Change Analysis of Pine Island Ice Shelf Based on Multi-Source Satellite Observations during 1973–2020

“…The frequency correlator of COSI-Corr estimates the phase difference in the Fourier domain and uses a bell-shaped function to avoid edge effects, and is less sensitive to outliers [31]. Fang et al [32] suggested that phase correlation in COSI-Corr with larger matching windows can achieve the best and most robust glacier motion estimation when applied to optical Landsat images. For fast-moving ice, Landsat image pairs with short repeat cycles (between 32 and 80 days) and ERS-1/2 image pairs with a 35-day repeat cycle are preferred.…”

Assessing the Variation of Glaciers Velocity in Hunza Basin, Gilgit Baltistan Using Advanced Geospatial Techniques

“…In recent decades, the optical image correlation technique for measuring displacement of features in earth and geological science has been widely used. In the feature tracking method, two images are correlated to measure displacement and one image is treated as a master image whereas the other is a slave [11,35]. In this technique, a matrix is selected on the master image with some center and the same window is dug out from the slave image and both are compared using similar function.…”

Analysis of the Velocity Changes of the Jakobshavn Glacier Based on SAR Imagery

“…Wherefore, the offset-tracking method is mainly used in ice-covered areas, where velocities are significant and major displacements appear in the horizontal plane. This technique is widely used to obtain glacier surface velocities (yan et al 2013, Strozzi et al 2017, Fang et al 2020, Neckel et al 2020 but also displacement values in mining areas (Huang et al 2016, 2020, Zhao et al 2021 or even in the areas of landslides (Cai et al 2017;Sun et al 2017) or sand dunes (Mahmoud et al 2020). The accuracy of this method is determined as 1/10-1/30 of the pixel size (Fan et al 2019).…”