Open-source feature-tracking algorithm for sea ice drift retrieval from Sentinel-1 SAR imagery

Stefan Muckenhuber,Stein Sandven,Anton Andreevich Korosov

doi:10.5194/tc-10-913-2016

Abstract

Abstract. A computationally efficient, open-source feature-tracking algorithm, called ORB, is adopted and tuned for sea ice drift retrieval from Sentinel-1 SAR (Synthetic Aperture Radar) images. The most suitable setting and parameter values have been found using four Sentinel-1 image pairs representative of sea ice conditions between Greenland and Severnaya Zemlya during winter and spring. The performance of the algorithm is compared to two other feature-tracking algorithms, namely SIFT (Scale-Invariant Feature Transform) and SURF (Speeded-Up Robust Features). Having been applied to 43 test image pairs acquired over Fram Strait and the north-east of Greenland, the tuned ORB (Oriented FAST and Rotated BRIEF) algorithm produces the highest number of vectors (177 513, SIFT: 43 260 and SURF: 25 113), while being computationally most efficient (66 s, SIFT: 182 s and SURF: 99 s per image pair using a 2.7 GHz processor with 8 GB memory). For validation purposes, 314 manually drawn vectors have been compared with the closest calculated vectors, and the resulting root mean square error of ice drift is 563 m. All test image pairs show a significantly better performance of the HV (horizontal transmit, vertical receive) channel due to higher informativeness. On average, around four times as many vectors have been found using HV polarization. All software requirements necessary for applying the presented feature-tracking algorithm are open source to ensure a free and easy implementation.

Highlights

Sea ice motion is an essential variable to observe from remote sensing data, because it strongly influences the distribution of sea ice on different spatial and temporal scales.Ice drift causes advection of ice from one region to another and export of ice from the Arctic Ocean to the sub-Arctic seas
Having been applied to 43 test image pairs acquired over Fram Strait and the north-east of Greenland, the tuned ORB (Oriented FAST and Rotated BRIEF) algorithm produces the highest number of vectors (177 513, Scale-Invariant Feature Transform (SIFT): 43 260 and Speeded-Up Robust Features (SURF): 25 113), while being computationally most efficient (66 s, SIFT: 182 s and SURF: 99 s per image pair using a 2.7 GHz processor with 8 GB memory)
CMEMS vs. manual 1690 0.98 −415 201 3440 ± 1105 culated drift results against manually derived vectors, we found that our algorithm (EORB = 563 m) had a distinctly higher accuracy than the drift data set provided by CMEMS (ECMEMS = 1690 m)

Summary

Introduction

Sea ice motion is an essential variable to observe from remote sensing data, because it strongly influences the distribution of sea ice on different spatial and temporal scales.Ice drift causes advection of ice from one region to another and export of ice from the Arctic Ocean to the sub-Arctic seas. There is still a lack of extensive sea ice drift data sets with sufficient resolution to estimate convergence and divergence on a spatial scaling of less than 5 km. The regions of interest are the ice-covered seas between Greenland and Severnaya Zemlya, i.e. the Greenland Sea, Barents Sea, Kara Sea and the adjacent part of the Arctic Ocean. This area is characterized by a strong seasonal cycle of sea ice cover, a large variation of different ice classes (multiyear ice, first-year ice, marginal ice zone etc.) and a wide range of drift speeds (e.g. strong ice drift in Fram Strait)

Objectives

Methods

Results

Conclusion