Sea Ice Images Research Articles

SAR Sea-Ice Image Analysis Based on Iterative Region Growing Using Semantics

Synthetic aperture radar (SAR) has been intensively used for sea-ice monitoring in polar regions. A computer-assisted analysis of SAR sea-ice imagery is extremely difficult due to numerous imaging parameters and environmental factors. This paper presents a system which, with some limited information provided, is able to perform an automated segmentation and classification for the SAR sea-ice imagery. In the system, both the segmentation and classification processes are based on a Markov random-field theory and are formulated in a joint manner under the Bayesian framework. Solutions to the formulation are obtained by a region-growing technique which keeps refining the segmentation and producing semantic class labels at the same time in an iterative manner. The algorithm is a general-segmentation approach named iterative region growing using semantics, which, in this paper, is dedicated to the problem of classifying the operational SAR sea-ice imagery provided by the Canadian Ice Service (CIS). The classified image results have been validated by the CIS personnel, and the resulting classifications are quite successful using the same algorithm applied to diverse data sets.

Multisensor approach to automated classification of sea ice image data

A multisensor data fusion algorithm based on a multilayer neural network is presented for sea ice classification in the winter period. The algorithm uses European Remote Sensing (ERS), RADARSAT synthetic aperture radar (SAR), and low-resolution television camera images and image texture features. Based on a set of in situ observations made at the Kara Sea, a neural network is trained, and its structure is optimized using a pruning method. The performance of the algorithm with different combinations of input features (sensors) is assessed and compared with the performance of a linear discriminant analysis (LDA)-based algorithm. We show that for both algorithms a substantial improvement can be gained by fusion of the three different types of data (91.2% for the neural network) as compared with single-source ERS (66.0%) and RADARSAT (70.7%) SAR image classification. Incorporation of texture increases classification accuracy. This positive effect of texture becomes weaker with increasing number of sensors (from 8.4 to 6.4 percent points for the use of two and three sensors, respectively). In view of the short training time and smaller number of adjustable parameters, this result suggests that semiparametric classification methods can be considered as a good alternative to the neural networks and traditional parametric statistical classifiers applied for the sea ice classification.

Unsupervised segmentation of synthetic aperture Radar sea ice imagery using a novel Markov random field model

Environmental and sensor challenges pose difficulties for the development of computer-assisted algorithms to segment synthetic aperture radar (SAR) sea ice imagery. In this research, in support of operational activities at the Canadian Ice Service, images containing visually separable classes of either ice and water or multiple ice classes are segmented. This work uses image intensity to discriminate ice from water and uses texture features to identify distinct ice types. In order to seamlessly combine image spatial relationships with various image features, a novel Bayesian segmentation approach is developed and applied. This new approach uses a function-based parameter to weight the two components in a Markov random field (MRF) model. The devised model allows for automatic estimation of MRF model parameters to produce accurate unsupervised segmentation results. Experiments demonstrate that the proposed algorithm is able to successfully segment various SAR sea ice images and achieve improvement over existing published methods including the standard MRF-based method, finite Gamma mixture model, and K-means clustering.

ARKTOS: An Intelligent System for SAR Sea Ice Image Classification

We present an intelligent system for satellite sea ice image analysis named Advanced Reasoning using Knowledge for Typing Of Sea ice (ARKTOS). ARKTOS performs fully automated analysis of synthetic aperture radar (SAR) sea ice images by mimicking the reasoning process of sea ice experts. ARKTOS automatically segments a SAR image of sea ice, generates descriptors for the segments of the image, and then uses expert system rules to classify these sea ice features. ARKTOS also utilizes multisource data fusion to improve classification and performs belief handling using Dempster-Shafer. As a software package, ARKTOS comprises components in image processing, rule-based classification, multisource data fusion, and graphical user interface-based knowledge engineering and modification. As a research project over the past ten years, ARKTOS has undergone phases such as knowledge acquisition, prototyping, refinement, evaluation, deployment, and operationalization at the U.S. National Ice Center. In this paper, we focus on the methodology, evaluations, and classification results of ARKTOS.

Towards multi-sensor microwave remote sensing of frost flowers on sea ice

AbstractIt has been hypothesized that frost flowers could be important for tropospheric chemistry and, as a source of sea-salt aerosol, for the interpretation of ice-core data. Furthermore, frost flowers can cause severe errors in sea-ice products derived from remote-sensing data. Up to now there have been few datasets available relating frost-flower distribution and variability in the sea-ice-covered regions. We present a method for frost-flower detection using a combination of active and passive microwave sensors. The vertical polarized C-band radar backscatter of a young sea-ice region covered with frost flowers can be higher than the signal from multi-year ice or from the wind-roughened ocean. Therefore, the classification result of synthetic aperture radar (SAR) sea-ice images can be defective due to the influence of frost flowers. The ambiguity of the open-water and frost-flower backscatter signal was successfully resolved using additional information from the 85 GHz Special Sensor Microwave/Imager (SSM/I) channels. Results of the method were compared to aircraft measurements and observations in the marginal ice zone near Svalbard.

Incidence angle dependence of the statistical properties of C-band HH-polarization backscattering signatures of the Baltic Sea ice

Incidence angle dependence of three statistical parameters-the mean of the backscattering coefficient (/spl sigma//spl deg/), standard deviation, and autocorrelation coefficient of texture (/spl sigma//sub T/ and /spl rho//sub T/)-of the C-band horizontal-horizontal (HH) polarization backscattering signatures of the Baltic Sea ice are investigated using RADARSAT ScanSAR Narrow images and helicopter-borne Helsinki University of Technology Scatterometer (HUTSCAT) data. The analysis of the large amount of data shows that the relationship between the mean /spl sigma//spl deg/ in decibel scale and the incidence angle in the range from 19/spl deg/ to 46/spl deg/ is usually well described by a linear model. In general, the RADARSAT and HUTSCAT results agree with each other, and they are also supported by theoretical backscattering model calculations; the more deformed the ice, the smaller the slope between /spl sigma//spl deg/ and the incidence angle, and the higher the moisture content of snow or ice, the larger the slope. The derived /spl sigma//spl deg/ incidence angle dependencies can be used to roughly compensate the /spl sigma//spl deg/ incidence angle variation in the SAR images to help their visual and automated classification. The variability of /spl sigma//sub T/ and /spl rho//sub T/ with the increasing incidence angle is insignificant compared to the variability within each ice type. Their average changes with the incidence angle are so small that, in practice, their trends do not need to be compensated. The results of this study can be utilized when developing classification algorithms for the RADARSAT ScanSAR and ENVISAT HH-polarization Wide Swath images of the Baltic Sea ice.

ARKTOS: a knowledge engineering software tool for images

The goal of our ARKTOS project is to build an intelligent knowledge-based system to classify satellite sea ice images. It involves acquiring knowledge from sea ice experts, quantifying such knowledge as computational entities and ultimately building an intelligent classifier. In this paper we describe a two-stage knowledge engineering approach that facilitates explicit knowledge transfer, converting implicit visual cues and cognition of the experts to explicit attributes and rules implemented by the engineers. First, there is a prototyping stage that involves interviewing sea ice experts, transcribing the sessions, identifying descriptors and rules, designing and implementing the knowledge and delivering the prototype. The objective of this stage is to obtain a modestly accurate classification system quickly. Second, there is a refinement stage that involves evaluating the prototype, refining the knowledge base, modifying the design and re-evaluating the improved system. Since the refinement is evaluation-driven, the experts and the engineers are motivated explicitly to improve the knowledge base and are able to communicate with each other using a common, consistent platform. Moreover, since the classification result is immediately available, both sides are able to efficiently assess the correctness of the system. To facilitate the knowledge engineering of the second stage, we have designed and built three Java-based graphical user interfaces: arktosGUI, arktosViewer and arktosEditor. arktosGUI concentrates on feature-based refinement of specific attributes and rules. arktosViewer deals with regional evaluation. arktosEditor has a rule indexing and search mechanism and knowledge base editing capabilities.

Surface roughness characterizations of sea ice and ice sheets: case studies with MISR data

This work is an examination of potential uses of multiangular remote sensing imagery for mapping and characterizing sea ice and ice sheet surfaces based on surface roughness properties. We use data from the Multi-angle Imaging SpectroRadiometer (MISR) to demonstrate that ice sheet and sea ice surfaces have characteristic angular signatures and that these angular signatures may be used in much the same way as spectral signatures are used in multispectral classification. Three case studies are examined: sea ice in the Beaufort Sea off the north coast of Alaska, the Jakobshavn Glacier on the western edge of the Greenland ice sheet, and a region in Antarctica south of McMurdo station containing glaciers and blue-ice areas. The MISR sea ice image appears to delineate different first-year ice types and, to some extent, the transition from first-year to multiyear ice. The MISR image shows good agreement with sea ice types that are evident in concurrent synthetic aperture radar (SAR) imagery and ice analysis charts from the National Ice Center. Over the Jakobshavn Glacier, surface roughness data from airborne laser altimeter transects correlate well with MISR-derived estimates of surface roughness. In Antarctica, ablation-related blue-ice areas, which are difficult to distinguish from bare ice exposed by crevasses, are easily detected using multiangular data.

Arctic sea-ice albedo derived from RGPS-based ice-thickness estimates

AbstractThe RADARSAT geophysical processor system (RGPS) uses sequential synthetic aperture radar images of Arctic sea ice taken every 3 days to track a large set of Lagrangian points over the winter and spring seasons. The points are the vertices of cells, which are initially square and 10 km on a side, and the changes in the area of these cells due to opening and closing of the ice are used to estimate the fractional area of a set of first-year ice categories. The thickness of each category is estimated by the RGPS from an empirical relationship between ice thickness and the freezing degree-days since the formation of the ice. With a parameterization of the albedo based on the ice thickness, the albedo may be estimated from the first-year ice distribution. We compute the albedo for the first spring processed by the RGPS, the early spring of 1997. The data include most of the Beaufort and Chukchi Seas. We find that the mean albedo is 0.79 with a standard deviation of 0.04, with lower albedo values near the edge of the perennial ice zone. The biggest source of error is likely the assumed rate of snow accumulation on new ice.

Determining the floe-size distribution of East Antarctic sea ice from digital aerial photographs

AbstractFloe size within the Antarctic pack ice is an important parameter that affects both the ocean-ice-atmosphere energy exchange and the mechanical properties of pack ice. In this paper we present a computer-based algorithm to extract floe-size distribution information from digital aerial photographs of Antarctic sea ice. The algorithm segments digital images of sea ice into distinct floes from which the size of each floe is calculated and floe-size distributions are derived. Through a recursive application of the morphological erosion operator, each floe is eroded towards its centre while each pixel’s erosion number is recorded. This ensures no information is lost while individual floes are identified. The algorithm combines image-processing techniques with some manual input to accurately identify the boundaries of individual floes within an image. To demonstrate the functionality of the algorithm, six images, representing regions along a 22 km south-to-north transect through a transitional zone of pack ice near 65° S, 140° E, have been processed. Regional variations in the floe-size distributions show an increase in the number of smaller floes relative to larger floes and a decrease in the dominant floe size (in terms of areal coverage) from south to north. These results are consistent with ship-based observations.

Separating touching objects in remote sensing imagery: the restricted growing concept and implementations

This paper defines the restricted growing concept (RGC) fur object separation and provides an algorithmic analysis of its implementations. Our concept decomposes the problem of object separation into two stages. First, separation is achieved by shrinking the objects to their cores while keeping track of their originals as masks. Then the core is grown within the masks obeying the guidelines of a restricted growing algorithm. In this paper, we apply RGC to the remote sensing domain, particularly the synthetic aperture radar (SAR) sea ice images.

Information states in radar imagery of sea ice

It is shown empirically that information in an image of sea ice exists in two forms-textural and structural. It is shown that such information can be extracted from the conditional probability of intensity differences between neighboring pixels. There exists a relationship between the two forms of information, called the information curve, which has identifiable subsections, each related to an ice type. The overall state of the sea ice image, composed as it is of a mixture of ice types, can therefore be given by the location of phase transition points that mark the end points of the subsections on the information curve composed of the two types of information. This result leads to a segmentation of the image by ice type. The properties of the information within the image are examined. It is shown that the textural information associated with individual ice pieces is scale invariant, whereas the structural information associated with the size and shape of these pieces, which dominates the image, has a scaling or fractal property. These results led to the examination of similarity sets, describing scaled intensity differences and separation jointly. Significantly, a multifractal relationship for these sets holds at all intensities, at least for ice types associated with brittle fracture. Another result inspired by the need to identify paths of least resistance through an ice field is the existence of a distinct change in the connectivity of these similarity sets at a critical value of the index parameter of the set. Such a critical value is in turn related to that of the local textural information, which is related to the phase transitions associated with the relationship between textural and structural information.

Unsupervised segmentation of ERS and Radarsat sea ice images using multiresolution peak detection and aggregated population equalization

In this paper we describe Automated Sea Ice Segmentation (ASIS), a system that automatically segments Synthetic Aperture Radar (SAR) sea ice imagery. This system integrates image processing, data mining, and machine learning methodologies to determine the number of visually separable classes in ERS and Radarsat sea ice images. We introduce two new techniques: multiresolution peak detection and spatial clustering. The detection is a noise-resistant data discretization methodology that results in an initial segmentation of the image. The clustering is based on an innovative concept called Aggregated Population Equalization that utilizes spatial relationships among classes to merge and split the population environment. Its self-organizing ability produces the final segmentation and automates ASIS. In addition, we have designed a Java-based graphical user interface that facilitates post-segmentation human evaluation and classification. Thus, ASIS can be used as a pre-processor to help analyse sea ice images as well as a basis for human classification of sea ice images. We have tested the system on more than 300 ERS-1, ERS-2 and Radarsat SAR sea ice images and analysed the results to point out the strengths and weaknesses of ASIS in the automated segmentation of sea ice images.

The Northern Sea Route, 1997

Abstract1997 was an important year for Russia's Northern Sea Route (NSR) in terms of foreign interest. The first full transits of the NSR by western merchant ships were made. The International Northern Sea Route Programme (INSROP) continued into phase 2 of its research projects, and planning proceeded for a European Union-sponsored demonstration voyage to the Ob' River. The nuclear icebreaker Arktika returned to Arctic operations, while two icebreakers were removed from polar service. Information about the NSR became more widely available with the distribution of an English translation of a navigation guide. Winter operations continued in the Kara Sea, albeit reduced, and a limited number of summer voyages to Russian Arctic ports originated from the Pacific rim. RADARS AT synthetic aperture radar images of sea ice along the NSR were used for the first time during several demonstration voyages.

Model-based interpretation of ERS-1 SAR images of Arctic sea ice

Interpretation of Synthetic Aperture Radar (SAR) images of sea ice is complex because of the natural variability of sea ice and sensor-induced effects, such as speckle. Most of the research on SAR image interpretation has focused on the winter months and algorithms were developed to classify sea ice successfully under cold conditions. However, interpretation of SAR images during the seasonal transitions has proved difficult due to rapidly changing weather conditions. In this paper we address the application of SAR during the transition from summer to the fall freeze-up. This period is important because it signals the start of significant new ice growth, which affects the air-ocean heat exchange and injects brine into the upper layers of the ocean. We have interpreted SAR images of the sea ice in terms of the basic ice characteristics by using shipborne radar measurements of sea ice during the freeze-up and models derived from these measurements. We have shown that the model-based approach is effective in interpreting SAR images during this seasonal transition. This work also provides the physical mechanisms responsible for the large increase in backscatter observed at the end of the summer melt season.

A laboratory study of the effect of frost flowers on C band radar backscatter from sea ice

C band images of Arctic sea ice taken by the ERS 1 synthetic aperture radar show transitory regions of enhanced radar backscatter from young sea ice. Published field observations associate this increase with frost flower growth and the capture of blowing snow by the flowers. To investigate the first part of this phenomenon, we carried out a laboratory experiment on the response of C band radar backscatter to frost flowers growing on the surface of newly formed saline ice. The experiment took place in a 5 m by 7 m by 1.2 m deep saline water pool located in a two‐story indoor refrigerated facility at the Cold Regions Research and Engineering Laboratory. Sodium chloride ice was grown in this pool at an air temperature of −28°C. The frost flowers first appeared on the ice surface as dendrites and then changed to needles as the ice sheet grew thicker and the surface temperatures became colder. The frost flowers reached to a height of 10–15 mm, and beneath each cluster of frost flowers a slush layer formed to a thickness of approximately 4 mm. Far‐field radar measurements of the backscatter from the ice were made at incident angles from 20° to 40° and at approximately 6‐hour intervals throughout the 3‐day period of the experiment. A backscatter minimum occurred early in the flower growth at the time coincident with an abrupt doubling in the ice surface salinity. Once the full flower coverage was achieved, we removed first the crystal flowers and then the slush layer from the ice surface. The results for these cases show that the crystals have little impact on the backscatter, while the underlying slush patches yield a backscatter increase of 3–5 dB over that of bare ice. The laboratory results suggest that this relative backscatter increase of approximately 5 dB can be used as an index to mark the full areal coverage of frost flowers.

Discrimination of sea ice in the Labrador marginal ice zone from synthetic aperture radar image texture

This study addresses the modelling of synthetic aperture radar (SAR) image texture for sea ice scenes in the Labrador marginal ice zone (MIZ). The image texture of distributed scatterers contains a substantial component relating to the imaging system as well as information about the scene itself. Theory shows that the image autocorrelation function (ACF) may be analysed to separate system contributions from scene contributions under certain conditions. The main theses of the study are: (i) SAR intensity images of sea ice are spatially nonGaussian; and (ii) the predominant types and forms of MIZ sea ice may be discriminated based upon ACF model parameters. Experimental results indicate that the model provides an excellent fit to the measured ACFs. The image texture was found to be a strong function of the form of the sea ice as well as its type. For a given type, the various forms could be discriminated through a single SAR channel. For full discrimination of all types and forms observed, a two-channel combination was necessary: XHV CHH or XHV CHV.

Speckle reduction and segmentation of Synthetic Aperture Radar images

Two simple modifications to the established sigma filter are proposed, and demonstrated to improve both its efficiency at smoothing speckle noise and its ability to preserve fine features. A fully automatic, unsupervised segmentation algorithm, based on the iterative application of the new filter, is described and successfully applied to ERS-1 Synthetic Aperture Radar (SAR) images of sea ice. The simplicity of this segmentation approach, together with its ability to segment classes separated by a gradual change in grey level intensity, may offer significant advantages over techniques based on edge-detection and region growing.

A finite mixtures algorithm for finding proportions in SAR images

This correspondence describes an algorithm for estimating the proportions of classes in an SAR image by first assuming that an image consists of a mixture of a known number of different pixel types. A maximum likelihood estimate of the parameters of the resulting mixture distribution is then evaluated using the EM algorithm. An advantage of the finite mixtures approach is that the quantities of interest, the proportions, are directly estimated. The technique is applied to aircraft synthetic aperture radar (SAR) images of sea ice. In addition to finding the proportions of the classes, knowledge of the mixture components allows image displays tailored to a user's requirements.

Segmentation of Synthetic-Aperture Radar imagery of sea ice

Abstract This paper describes a ‘segmentation’ algorithm, which has been designed for Synthetic-Aperture Radar (SAR) sea-ice images as part of the U.K. Earth Observation Data Centre (EODC) Science Team activity. The segmented sea-ice product consists of distinct, statistically-homogeneous regions of ice and water from which various atttributes can be determined, such as area, centroid position, mean intensity and standard deviation, shape, orientation and texture. Such attributes are important in providing a template from which it is possible to derive information such as ice type, concentration and dynamics, the position of the ice edge, and the characteristics of leads and polynyas. Although the method has been developed for processing ERS-1 data, it is demonstrated in this work using both simulations and SAR sea-ice data obtained from the Seasat campaign of 1978, as ERS-1 data were not yet available at the t ime of implementation.