Using a Phase-Congruency-Based Detector for Glacial Lake Segmentation in High-Temporal Resolution Sentinel-1A/1B Data

Abstract

Glacial lakes are the main products of glacier melting caused by climate warming. Changes in glacial lake outlines and in textural features in the surroundings are important indicators used in the dynamic monitoring of glacial lakes. Synthetic aperture radar (SAR) data, which can be used regardless of weather and solar illumination conditions, are a useful source of information about the occurrence and evolution of such lakes. The development of high-temporal resolution SAR makes it possible to study glacial lakes more frequently. However, segmentation of the glacial lakes using SAR data remains a challenge because of noise effects and low contrast. This paper presents a phase-congruency-based detector that can be used to extract the linear features (outlines and textural features) of glacial lakes using C-band high-temporal resolution Sentinel-1A/1B imagery of the Himalayas. In this algorithm, all the weak edges are first detected by phase congruency, which is insensitive to image contrast and magnification. Then, the fast-marching method is applied to improve the connectivity and form a complete lake boundary. Experimental results derived from intensity and coherence images are evaluated and compared for different lake types. It is shown that the proposed method can locate the exact position of most of the outlines. The relative errors for all the lakes in the intensity images were reduced to within four pixels with a 90% confidence level, and within three pixels for unconnected glacial lakes in the coherence data.