Automatic Detection of Small Icebergs in Fast Ice Using Satellite Wide-Swath SAR Images

Soldal, Ingri Halland; Dierking, Wolfgang; Korosov, Anton; Marino, Armando

doi:10.3390/rs11070806

Cited by 27 publications

(21 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It could be argued that too many objects are not labelled, leaving them out of training, but in opposition to this, one could say that too many small objects are included, and these are not of great importance. In Figure 11, training objects are also seen to be located within the large piece of floating ice in the right side of the image, this also raises the question of icebergs being present in other types of ice, such as in this study [23], or if such large pieces of floating ice should be in a class of their own or maybe not be included at all. Since ships are clearly defined objects, populating a SAR dataset with these does not face the same issues as the icebergs.…”

Section: Discussionmentioning

confidence: 71%

Deep Learning for Detecting and Classifying Ocean Objects: Application of YoloV3 for Iceberg–Ship Discrimination

Hass

Arsanjani

2020

IJGI

View full text Add to dashboard Cite

Synthetic aperture radar (SAR) plays a remarkable role in ocean surveillance, with capabilities of detecting oil spills, icebergs, and marine traffic both at daytime and at night, regardless of clouds and extreme weather conditions. The detection of ocean objects using SAR relies on well-established methods, mostly adaptive thresholding algorithms. In most waters, the dominant ocean objects are ships, whereas in arctic waters the vast majority of objects are icebergs drifting in the ocean and can be mistaken for ships in terms of navigation and ocean surveillance. Since these objects can look very much alike in SAR images, the determination of what objects actually are still relies on manual detection and human interpretation. With the increasing interest in the arctic regions for marine transportation, it is crucial to develop novel approaches for automatic monitoring of the traffic in these waters with satellite data. Hence, this study aims at proposing a deep learning model based on YoloV3 for discriminating icebergs and ships, which could be used for mapping ocean objects ahead of a journey. Using dual-polarization Sentinel-1 data, we pilot-tested our approach on a case study in Greenland. Our findings reveal that our approach is capable of training a deep learning model with reliable detection accuracy. Our methodical approach along with the choice of data and classifiers can be of great importance to climate change researchers, shipping industries and biodiversity analysts. The main difficulties were faced in the creation of training data in the Arctic waters and we concluded that future work must focus on issues regarding training data.

show abstract

Section: Discussionmentioning

confidence: 71%

Deep Learning for Detecting and Classifying Ocean Objects: Application of YoloV3 for Iceberg–Ship Discrimination

Hass

Arsanjani

2020

IJGI

View full text Add to dashboard Cite

show abstract

“…Figure 1 presents an example of the spatial coverages of the EW data acquired by S1A and S1B within six days in 2019. Additionally, the EW and IW data acquired in the Arctic are generally in polarization combination of co‐polarization and cross‐polarization, dedicated for sea ice monitoring (e.g., Hong & Yang, 2018; Li, Sun, et al., 2020; Soldal et al., 2019). With a spatial resolution of 40 m, S1 EW images can generally yield good observations of ocean waves, as illustrated in Figure 2.…”

Section: Introductionmentioning

confidence: 99%

Retrieval of Ocean Wave Heights From Spaceborne SAR in the Arctic Ocean With a Neural Network

Huang

2021

JGR Oceans

View full text Add to dashboard Cite

Prior to the launch of the Chinese French Oceanic Satellite (CFOSAT) with its onboard Surface Waves Investigation and Monitoring (SWIM) sensor, the only sensor capable of imaging ocean waves in two dimensions from space was the spaceborne synthetic aperture radar (SAR), which provides images with high spatial resolution. The SAR imaging mechanism of ocean waves is complex which is generally explained by three modulations: tilt modulation, hydrodynamic modulation and velocity bunching (Alpers et al., 1981; Valenzuela, 1978). While tilt and hydrodynamic modulations are also shared by real-aperture radar as the dominant imaging mechanisms of ocean waves, velocity bunching is unique for SAR to image ocean waves. The moving scatterer of water particles with a velocity either toward or away from a moving SAR sensor, causes an azimuthal shift in SAR images. In addition, velocity bunching in the SAR resolution cell leads to an azimuth cutoff , that is, the minimum SAR-detectable wavelength of ocean waves traveling in the azimuth direction. Therefore, the nonlinearity of SAR ocean wave imaging complicates their retrieval. In the following, we briefly summarize the existing methods used to retrieve ocean wave information in terms of both two-dimensional spectra and integral wave parameters. The Max Planck Institute (MPI) scheme developed by Hasselmann and Hasselmann (1991) and Hasselmann et al. (1996) is a widely used method to retrieve two-dimensional ocean wave spectra from spaceborne

show abstract

“…The complex nature of icebergs and sea ice unfortunately does not allow for the discrimination of icebergs with respect to other large sea ice features such as hummocks. A new detector (iDPolRaD) devised by Marino et al [13] has tried to address the problem of detecting small icebergs (<120 m) embedded within sea ice [19].…”

Section: Introductionmentioning

confidence: 99%

“…However, detectors can still identify icebergs at sizes of approximately 1-4 pixels. Several papers [19,27,28] highlight the ability to detect icebergs of medium to large size (120 m or larger). Nevertheless, as pointed out in Soldal et al [19], work is still required to improve the detection ability for icebergs smaller than 120 m. These icebergs are mostly calving off glacier tongues and ice sheets in the Arctic, and their smaller masses can go undetected due to factors such as the sea state and meteorological conditions [29,30].…”

Section: Introductionmentioning

confidence: 99%

Comparison of Target Detectors to Identify Icebergs in Quad-Polarimetric L-Band Synthetic Aperture Radar Data

Bailey¹,

Marino²,

Akbari³

2021

Remote Sensing

Self Cite

View full text Add to dashboard Cite

Icebergs represent hazards to ships and maritime activities and therefore their detection is essential. Synthetic Aperture Radar (SAR) satellites are very useful for this, due to their capability to acquire data under cloud cover and during day and night passes. In this work, we compared six state-of-the-art polarimetric target detectors to test their performance and ability to detect small-sized icebergs <120 m in four locations in Greenland. We used four single-look complex (SLC) ALOS-2 quad-polarimetric images from JAXA for quad-polarimetric detection and we compared with dual-polarimetric detectors using only the channels HH and HV. We also compared these detectors with single-polarimetric intensity channels and we tested using two scenarios: open ocean and sea ice. Our results show that the multi-look polarimetric whitening filter (MPWF) and the optimal polarimetric detector (OPD) provide the most optimal performance in quad- and dual-polarimetric mode detection. The analysis shows that, overall, quad-polarimetric detectors provide the best detection performance. When the false alarm rate (PF) is fixed to 10−5, the probabilities of detection (PD) are 0.99 in open ocean and 0.90 in sea ice. Dual-polarimetric or single-polarimetric detectors show an overall reduction in performance (the ROC curves show a decrease), but this degradation is not very large (<0.1) when the value of false alarms is relatively high (i.e., we are interested in bigger icebergs with a brighter backscattering >120 m, as they are easier to detect). However, the differences between quad- and dual- or single-polarimetric detectors became much more evident when the PF value was fixed to low detection probabilities 10−6 (i.e., smaller icebergs). In the single-polarimetric mode, the HV channel showed PD values of 0.62 for open ocean and 0.26 for sea ice, compared to values of 0.81 (open ocean) and 0.77 (sea ice) obtained with quad-polarimetric detectors.

show abstract

Automatic Detection of Small Icebergs in Fast Ice Using Satellite Wide-Swath SAR Images

Cited by 27 publications

References 18 publications

Deep Learning for Detecting and Classifying Ocean Objects: Application of YoloV3 for Iceberg–Ship Discrimination

Deep Learning for Detecting and Classifying Ocean Objects: Application of YoloV3 for Iceberg–Ship Discrimination

Retrieval of Ocean Wave Heights From Spaceborne SAR in the Arctic Ocean With a Neural Network

Comparison of Target Detectors to Identify Icebergs in Quad-Polarimetric L-Band Synthetic Aperture Radar Data

Contact Info

Product

Resources

About