Automatic Sea-Ice Classification of SAR Images Based on Spatial and Temporal Features Learning

Song, Wei; Li, Minghui; Gao, Wen; Huang, Dongmei; Ma, Zhenling; Liotta, Antonio; Perra, Cristian

doi:10.1109/tgrs.2020.3049031

Cited by 28 publications

(23 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To maintain the position consistency of the two satellite images, we cut out the non-overlapping regions of the data in the study. Following our previous study [31], we selected Hudson Bay as the study area, which locates in the central-eastern part of Canada and its sea ice condition has been long-term monitored by Canadian Ice Service (CIS). It has a large area and a long freezing process, which can provide a large sea ice coverage and different sea ice types and is suitable for data preparation.…”

Section: Sar Source and Study Areamentioning

confidence: 99%

“…The steps are shown in Figure 3. The preprocessing steps in this paper are a more comprehensive preprocessing scheme than our previous study [31], adding the reduction of the border noise and speckle noise of the SAR image, as well as the mitigation of the incidence angle dependence. It should be noted that the preprocessing does not really change the SAR image resolution.…”

Section: Sar Image Preprocessingmentioning

confidence: 99%

“…To produce representative data samples for training and testing sea ice classification models, SAR images need to be filtered to meet the sample requirements. In addition to the sea ice concentration proposed in paper [31], factors such as the development process of sea ice, the interswath discontinuities effect, the scalloping effect on the first subswath, as well as the edge mismatch between the SAR image and the ice chart will affect the quality of the dataset. To alleviate the influence of these factors, we refer to the ice chart and ice data that have the least time difference between SAR image acquisition time and ice charting time, and select sample scenes and regions according to the following explained principles:…”

Section: Sample Productionmentioning

confidence: 99%

“…It is clear that taking temporal information into consideration is helpful for sea ice studies. In addition, we previously studied the impact of the combined use of spatial and temporal information on sea ice classification, and the results show that the addition of temporal information is helpful for sea ice classification [31].…”

Section: Introductionmentioning

confidence: 99%

“…Each sample is a sequence of SAR image patches with six time-steps (about six weeks). The time-step of 6 is determined based on our previous study [31]. Nevertheless, this dataset can be used with any time-steps between one to six depending on the users' needs, simply by removing the unwanted image patches from the sequence.…”

mentioning

confidence: 99%

See 4 more Smart Citations

SI-STSAR-7: A Large SAR Images Dataset with Spatial and Temporal Information for Classification of Winter Sea Ice in Hudson Bay

Song

Gao

et al. 2021

Remote Sensing

Self Cite

View full text Add to dashboard Cite

Remote sensing satellites have been broadly applied to sea ice monitoring. The substantial increase in satellite imagery provides a large amount of data support for deep learning methods in the sea ice classification field. However, there is a lack of public remote sensing datasets to facilitate sea ice classification with spatial and temporal information and to benchmark the deep learning methods. In this paper, we provide a labeled large sea ice dataset derived from time-series sentinel-1 SAR images, dubbed SI-STSAR-7, and a validated dataset construction method for sea ice classification research. The SI-STSAR-7 dataset includes seven different sea ice types corresponding to different sea ice development stages in Hudson Bay during winter, and its samples are time sequences of SAR image patches in order to embody the differences of backscattering intensity and textures between different sea ice types, as well as the change of sea ice with time. We construct the dataset by first performing noise reduction and mitigation of incidence angle dependence on SAR images, and then producing data samples and labeling them based on our proposed sample-producing principles and the weekly regional ice charts provided by Canadian Ice Service. Three baseline classification methods are developed on SI-STSAR-7 to establish benchmarks, which are evaluated with accuracy and kappa coefficient. The sample-producing principles are verified through experiments. Based on the experimental results, sea ice classification can be implemented well on SI-STSAR-7.

show abstract

Section: Sar Source and Study Areamentioning

confidence: 99%

Section: Sar Image Preprocessingmentioning

confidence: 99%

Section: Sample Productionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations