Sea Ice Classification via Deep Neural Network Semantic Segmentation

Dowden, Benjamin; Silva, Oscar De; Huang, Weimin; Oldford, Dan

doi:10.1109/jsen.2020.3031475

Cited by 31 publications

(12 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The added granularity provided by our solution allows a better treatment of important phenomena such as the formation of fractures and leads that can substantially alter the structure of sea ice as it allows more short-wavelength absorption by the ocean [48]. Additionally, since tasked high-resolution satellite imagery (e.g., WV-3) can be retrieved at specified locations within hours, our approach can enhance sea ice detection for shipping and logistics with a broader range of action than ship-based camera approaches (e.g., [24,25]). Because of its reliability outside of pack-ice areas (e.g., Figure 5), our pipeline is capable not only of producing sharp ice floe segmentation masks but detecting the presence of floes in very-high resolution imagery.…”

Section: Discussionmentioning

confidence: 99%

“…With the concomitant popularization of high-resolution sensors, DL solutions have largely replaced methods such as Support Vector Machines (SVM) and has already become a staple in some areas of remote sensing [23]. In contrast to other works that use DL for classifying sea ice at medium resolution (e.g., [20,21]) and segment out sea ice in ship-borne images [24,25], the goal of the present work is extracting precise ice floe masks from high-resolution imagery. More specifically, we are targeting ice floes only-a daunting task given the large number of potentially confounding fine-scale structures (e.g., slush, melt ponds, etc.)…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fine-Scale Sea Ice Segmentation for High-Resolution Satellite Imagery with Weakly-Supervised CNNs

Gonçalves

Lynch

2021

Remote Sensing

View full text Add to dashboard Cite

Fine-scale sea ice conditions are key to our efforts to understand and model climate change. We propose the first deep learning pipeline to extract fine-scale sea ice layers from high-resolution satellite imagery (Worldview-3). Extracting sea ice from imagery is often challenging due to the potentially complex texture from older ice floes (i.e., floating chunks of sea ice) and surrounding slush ice, making ice floes less distinctive from the surrounding water. We propose a pipeline using a U-Net variant with a Resnet encoder to retrieve ice floe pixel masks from very-high-resolution multispectral satellite imagery. Even with a modest-sized hand-labeled training set and the most basic hyperparameter choices, our CNN-based approach attains an out-of-sample F1 score of 0.698–a nearly 60% improvement when compared to a watershed segmentation baseline. We then supplement our training set with a much larger sample of images weak-labeled by a watershed segmentation algorithm. To ensure watershed derived pack-ice masks were a good representation of the underlying images, we created a synthetic version for each weak-labeled image, where areas outside the mask are replaced by open water scenery. Adding our synthetic image dataset, obtained at minimal effort when compared with hand-labeling, further improves the out-of-sample F1 score to 0.734. Finally, we use an ensemble of four test metrics and evaluated after mosaicing outputs for entire scenes to mimic production setting during model selection, reaching an out-of-sample F1 score of 0.753. Our fully-automated pipeline is capable of detecting, monitoring, and segmenting ice floes at a very fine level of detail, and provides a roadmap for other use-cases where partial results can be obtained with threshold-based methods but a context-robust segmentation pipeline is desired.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Fine-Scale Sea Ice Segmentation for High-Resolution Satellite Imagery with Weakly-Supervised CNNs

Gonçalves

Lynch

2021

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…NNs have attracted considerable research attention recently as a promising tool to avail automated sea-ice monitoring solutions [10]- [14], [20]. In [10], a NN algorithm was designed to classify sea-ice in SAR images of central arctic.…”

Section: A Related Workmentioning

confidence: 99%

“…The authors utilized video streams acquired by a webcam to generate the data set with nomenclature classes of water, ice, and clutter. In [20], two data sets were used to train DNNs. The scene in the first data set captures four classes, namely ice, vessel, ocean, and sky; while the scene in the second data set captures more ice classes.…”

Section: A Related Workmentioning

confidence: 99%

Improved Sea-Ice Identification Using Semantic Segmentation With Raindrop Removal

et al. 2022

Self Cite

View full text Add to dashboard Cite

Sea-ice identification is an essential process for safety critical navigation support of surface vessels in polar waters. Semantic segmentation has drawn much attention as an enabling technique for fast detection of objects in a scene including sea-ice conditions. Identifying sea-ice is a challenging problem, especially in the presence of raindrops. The raindrop alters the boundaries of the objects in the scene, and thus, degrades the identification performance. In this work, a raindrop removing framework is developed to enhance the classification performance. Three deep-learning semantic segmentation networks are trained to classify the scene of sea-ice images into ice, water, ship, and sky. The deep-learning networks are VGG-16, fully convolutional network, and pyramid scene parsing network. Transfer learning along with data augmentation operations have been implemented to improve the training process. Results illustrate that data augmentation operations enhance the performance of the three models. Moreover, the raindrop removing framework improves the models' performance, e.g. the average intersection over union of the VGG-16 model is improved from 85.91% to 91.70%.

show abstract

“…Previous attempts at automated analysis of ice scenes (closerange images) are either limited to a few freshwater ice features to understand/monitor river ice processes [30]- [33] or use traditional image processing techniques that are limited to broken ice with inclusions of brash, young gray, and frazil/nilas ice and specific lighting conditions [34]- [36]. A few recent studies have applied deep learning-based methods for the analysis of generalized ice scenes, but their focus is on the classification of the ice objects present in the optical image [37], [38] or segmentation of first-year ice types rather than on differentiation between ice objects (deformed ice, level ice, icebergs) [39]. Moreover, the challenges related to postprocessing ice object localizations with size-sensitive definitions, such as brash ice and ice floes, as well as the sensitivity of deep learning models to ice image distortions (e.g., grayscale and vignette effects), have never been explored before and are a subject of this paper.…”

Section: Relevant Workmentioning

confidence: 99%

Supplementing Remote Sensing of Ice: Deep Learning-Based Image Segmentation System for Automatic Detection and Localization of Sea-ice Formations From Close-Range Optical Images

2021

View full text Add to dashboard Cite

Sea Ice Classification via Deep Neural Network Semantic Segmentation

Cited by 31 publications

References 17 publications

Fine-Scale Sea Ice Segmentation for High-Resolution Satellite Imagery with Weakly-Supervised CNNs

Fine-Scale Sea Ice Segmentation for High-Resolution Satellite Imagery with Weakly-Supervised CNNs

Improved Sea-Ice Identification Using Semantic Segmentation With Raindrop Removal

Supplementing Remote Sensing of Ice: Deep Learning-Based Image Segmentation System for Automatic Detection and Localization of Sea-ice Formations From Close-Range Optical Images

Contact Info

Product

Resources

About