Baltic Sea ice SAR segmentation and classification using modified pulse-coupled neural networks

A method for segmentation and classification of Baltic Sea ice synthetic aperture radar (SAR) images, based on pulse-coupled neural networks (PCNNs), is presented. Also, automated training, which is based on decomposing the total pixel value distribution into a mixture of class distributions, is presented and discussed. The algorithms have been trained and tested using logarithmic scale Radarsat-1 ScanSAR Wide mode images over the Baltic Sea ice. Before the decomposition into mixture of class distributions, an incidence angle correction, specifically designed for these Baltic Sea ice SAR images, is applied. Because the data distributions in the uniform areas of these images are very close to Gaussian distributions, the data are decomposed into a mixture of Gaussian distributions, using the Expectation-Maximazation algorithm. Only uniform image areas are used in the decomposition phase. The mixture of distributions is compared to the distributions of the Baltic Sea ice classes, based on earlier scatterometer measurements and visual video interpretations of the sea ice classes. The parameter values for the PCNN segmentation are defined based on this mixture of distributions. The PCNN segmentation results are also compared to the operational sea ice information of digitized ice charts and to visual interpretation of the sea ice class.     

SAR image statistics related to atmospheric drag over sea ice

The possibility of using synthetic aperture radar (SAR) data to distinguish sea-ice regions with different atmospheric drag is explored. Both the amplitude of the radar return and statistics derived from SAR image data are examined. Roughness statistics data from several pack-ice areas are used in a backscatter model to predict the return from surfaces with measured drag coefficients. The results suggest that the scattering coefficient for typical radar wavelengths is insensitive to the roughness elements responsible for the observed drag coefficient variations over pack ice free of major ridges. For marginal ice zones, where ice concentration and floe deformation contribute to atmospheric drag, a simple model for the atmospheric boundary layer is used to provide qualitative relationships between drag coefficient and regional ice properties (ice concentration, floe size distribution, floe edge density) derivable from SAR data. Simple algorithms to produce maps of ice concentration and edge density are outlined and applied to 23.5-cm SAR digital image data     

Application of neural networks for sea ice classification inpolarimetric SAR images

Several automatic methods have been developed to classify sea ice types from fully polarimetric synthetic aperture radar (SAR) images, and these techniques are generally grouped into supervised and unsupervised approaches. In previous work, supervised methods have been shown to yield higher accuracy than unsupervised techniques, but suffer from the need for human interaction to determine classes and training regions. In contrast, unsupervised methods determine classes automatically, but generally show limited ability to accurately divide terrain into natural classes. In this paper, a new classification technique is applied to determine sea ice types in polarimetric and multifrequency SAR images, utilizing an unsupervised neural network to provide automatic classification, and employing an iterative algorithm to improve the performance. The learning vector quantization (LVQ) is first applied to the unsupervised classification of SAR images, and the results are compared with those of a conventional technique, the migrating means method. Results show that LVQ outperforms the migrating means method, but performance is still poor. An iterative algorithm is then applied where the SAR image is reclassified using the maximum likelihood (ML) classifier. It is shown that this algorithm converges, and significantly improves classification accuracy. The new algorithm successfully identifies first-year and multiyear sea ice regions in the images at three frequencies. The results show that L- and P-band images have similar characteristics, while the C-band image is substantially different. Classification based on single features is also carried out using LVQ and the iterative ML method. It is found that the fully polarimetric classification provides a higher accuracy than those based on a single feature. The significance of multilook classification is demonstrated by comparing the results obtained using four-look and single-look classifications     

C-band SAR observations of marginal ice zone rheology in theLabrador Sea

Synthetic-aperture radar (SAR) imagery collected during the LIMEX'87 field program in the Labrador Sea is used to qualitatively interpret the deformation modes in play throughout the experiment. The ice cover exhibited two distinct rheologies separated by a clear line of shear: a nonlinear viscous outer regime, and a quasi-brittle inner regime. A single constitutive relation capable of effectively modeling both is unlikely within a plastic rate-independent formulation. Rate-dependent effects are discussed in relation to favoring brittle fraction in an otherwise ductile material     

Sea ice identification using dual-polarized Ku-band scatterometerdata

This paper describes a classification algorithm using dual-polarized scatterometer measurements to identify the edge of the sea ice cover. The distinct polarization scattering signatures of sea ice and open water are discussed and illustrated with the dual-polarized radar measurements from the Seasat-A scatterometer (SASS). The analysis of SASS data suggests that the ratio of vertical and horizontal polarization backscatter, denoted as the copol ratio, is a useful discriminator of sea ice and open ocean. A simple classification algorithm using the thresholds of the copol ratio and backscatter levels is proposed. The feasibility of this algorithm is demonstrated using the SASS data from the single-sided, dual-polarization mode. The results indicate that the dual-polarized measurements from the NASA scatterometer (NSCAT) can be used to produce routine maps of sea ice edges     

利用灰度和纹理特征的SAR图像分类研究

多类别多特征量情况下的合成孔径雷达(SAR)图像的目标分类是一个难以解决的问题.从灰度和纹理模型出发,提出了综合利用灰度和纹理特征的目标分类方法.均值和方差是灰度模型中重要的特征统计量,而能量、熵、对比度、局部相似性和相关性是纹理模型中重要的特征统计量.灰度和纹理特征能确切地描述SAR图像中的目标.通过构造特征向量,定义向量之间的距离,并按照最小距离方法进行目标分类.以一定大小的窗口读入样本,提高了算法的运行速度和抗噪能力.理论上,窗口越大,特征向量值越接近真实值.窗口越小,边缘的分类精度越高.实验表明该方法较好地处理了多类别多特征量情况下的SAR图像分类问题,分类结果是有效的,这为SAR图像目标分类提供了一条简单可行的途径.     

LIMEX '87 ice surface characteristics: implications forC-band SAR backscatter signatures

Ice surface characterization data collected in 1987, during the Labrador Ice Margin Experiment, are analyzed to estimate the changes in snow and ice properties at the onset of melt. Surface measurements were made from an ice research vessel on several days (some of which had coincident remote sensing flights) at a number of locations in the marginal ice zone. These data are used as input parameters in a simple scattering model to simulate the effects of variations in material properties upon C-band scattering signatures     

Sea ice concentration, ice temperature, and snow depth using AMSR-E data

A summary of the theoretical basis and initial performance of the algorithms that are used to derive sea ice concentration, ice temperature, and snow depth on sea ice from newly acquired Earth Observing System-Aqua/Advanced Microwave Scanning Radiometer-EOS (AMSR-E) radiances is presented. The algorithms have been developed and tested using historical satellite passive microwave data and are expected to provide more accurate products, since they are designed to take advantage of the wider range of frequencies and higher spatial resolution of the AMSR-E microwave instrument. Validation programs involving coordinated satellite, aircraft, and surface measurements to determine the accuracies of these sea ice products and to improve further our capability to monitor global sea ice are currently underway.     

Large-scale inverse Ku-band backscatter modeling of sea ice

Polar sea ice characteristics provide important inputs to models of several geophysical processes. Microwave scatterometers are ideal for monitoring these regions due to their sensitivity to ice properties and insensitivity to atmospheric distortions. Many forward electromagnetic scattering models have been proposed to predict the normalized radar cross section (/spl sigma//spl deg/) from sea ice characteristics. These models are based on very small scale ice features and generally assume that the region of interest is spatially homogeneous. Unfortunately, spaceborne scatterometer footprints are very large (5-50 km) and usually contain very heterogeneous mixtures of sea ice surface parameters. In this paper, we use scatterometer data in a large-scale inverse modeling experiment. Given the limited data resolution, we adopt a simple geometric optics forward-scattering model to analyze surface and volume scattering contributions to observed Ku-band signatures. A model inversion technique based on recursive optimization of an objective function is developed. The result is a least squares estimate of three surface parameters: the power reflection coefficient at nadir, the rms surface slope, and the volume scattering albedo. Simulations demonstrate the performance of the method in the presence of noise. The inverse model is implemented using Ku-band image reconstructed data collected by the National Aeronautics and Space Administration scatterometer. The results are used to analyze and interpret /spl sigma//spl deg/ phenomena occurring in the Antarctic and the Arctic.     

Springtime C-band SAR backscatter signatures of LabradorSea marginal ice: measurements versus modeling predictions

During the March 1987 Labrador Ice Margin Experiment (LIMEX '87) two independent investigations were conducted to determine the C-band backscattering cross section of the marginal pack ice along the Newfoundland coast. In one experiment, data from a recently calibrated C-band airborne scatterometer were combined with C-band synthetic aperture radar (SAR) data to measure the normalized scattering cross section of the ice at incidence angles from 10° to 74° to within ±2 dB. In the other experiment, detailed measurements of ice surface roughness and surface properties were made and the radar cross sections were predicted from a scattering model. In the present study, measured and model results are combined and shown to be fully compatible. By extension, the results are expected to apply to any rubbled sea-ice surface when surface scattering dominates     

Terrain influences in SAR backscatter and attempts to theircorrection

Synthetic aperture radar (SAR) images reveal radiometric image distortions that are caused by terrain undulations. The authors present the results of a study extracting and investigating the various components of these terrain influences. An imaging model, is set up for the geometric rectification of the SAR image and for a reconstruction of the imaging geometry. A prerequisite for the setup of this model is the use of a digital elevation model. Eight different geometric parameters are derived and investigated for their influence on grey-value variations in the geocoded SAR image. Image grey-value variations of three major land-use classes-forest, agricultural land, and urban/suburban areas-are examined. Empirical models of the SAR-backscatter variations are used to describe the relations between image grey values and various geometric parameters     

Terrain classification in SAR images using principal componentsanalysis and neural networks

The development of a neural-network-based classifier for classifying three distinct scenes (urban, park, and water) from several polarized SAR images of the San Francisco Bay area is discussed. The principal components (PC) scheme or Karhunen-Loeve transform is used to extract the salient features of the input data, and to reduce the dimensionality of the feature space prior to the application to the neural networks. Using the PC scheme along with the polarized images used in the present study led to substantial improvements in the classification rates when compared with previous studies. When a combined polarization architecture was used, the classification rate for water, urban, and park areas improved to 100%, 98.7%, and 96.1%, respectively     

Non-Gaussian signal statistics in ocean SAR imagery

The authors have studied the significance of non-Gaussian signal statistics in some synthetic aperture radar (SAR) images of the ocean surface. The study consisted of calculating the amplitude histogram of the returned echoes from the images and comparing these with the Rayleigh- and Kν-distributions, corresponding to the Gaussian and non-Gaussian statistics, respectively. The images used were some C-band SAR data from the Canadian airborne SAR collected during the NORCSEX'88 campaign and some ERS-1 data collected during the NORCSEX'91 campaign. The analysis of the NORCSEX'88 data included studies of the dependency of the signal statistics on incidence angle and meteorological and imaging conditions. It was found, specifically at small incidence angles, that there was a significant deviation from Gaussian statistics. It was also found that when the wind was blowing against the waves, the deviation from Gaussian statistics was more pronounced than when the wind was blowing in the same direction as the waves were propagating. The study also showed a correlation between the signal statistics and the width of the SAR image spectra. At low incidence angles, this agrees with the interpretation that non-Gaussian statistics may be related to strong widebanded scattering events. However, since non-Gaussian statistics also were observed at incidence angles as high as 50°, it is evident that the modulation of the scattering cross section by the long waves is also an important factor. In addition, the analysis of the ERS-1 data showed that to account for the width of the SAR image spectra, an azimuth smearing term, due to short scene coherence time, had to be included. This was in the present work done by modeling the short-coherence-time-smearing as a Gaussian low-pass filter. By this procedure, the authors were able to obtain realistic estimates for the average scene coherence time of the SAR scenes     

Satellite altimetry using ocean backscatter

A model for the signal received when satellite altimetry pulses are scattered back by the ocean's surface is postulated and analyzed. The second-order statistics of the complex nonstationary random process appearing at the detector input are graphically and analytically displayed as a function of the sea state. The standard deviation of various altitude estimates derived by subsequent operations on the detector output are computed, too. The analysis applies both to ordinary short-pulse altimeters and to systems that employ pulse compression.     

Terrain influences on SAR backscatter around Mt. Taranaki, NewZealand

An L-band SAR image from the Japanese JERS-1 satellite has been analyzed for the effects of local surface orientation relative to the radar illumination direction. Orthorectification of the SAR imagery and determination of the local surface orientation is achieved with the aid of a high resolution digital terrain model. An improved method of determining scatterer density for a surface in three dimensions is introduced and used to correct radiometrically the image for terrain variation. Residual radiometric effects due to surface orientation are shown to be dependent on the ground cover class. Backscatter from the indigenous forest was more isotropic than that from the farmland. As accurate registration was required for this study, a method for identifying control points in the rectified imagery is described which alleviated the difficulty of identifying them in the raw image     

Mixed pixel classification with robust statistics

The authors present a novel method for mixed pixel classification where the Hough transform and the trimmed means methods are used to classify small sets of pixels. They compare the performance of these methods with the least squares error method, and they show that in the presence of outliers, the trimmed means method is far more reliable than the traditional least squares error method, and even when no outliers are present, its performance is comparable to that of the least squares error method. The method is exhaustively tested using simulated data, and it is also applied to real Landsat TM data for which ground data are available     

Unvoiced/voiced classification and voiced harmonic parameters estimation using the third-order statistics

Unvoiced/voiced classification of speech is a challenging problem especially under conditions of low signal-to-noise ratio or the non-white-stationary noise environment. To solve this problem, an algorithm for speech classification, and a technique for the estimation of pairwise magnitude frequency in voiced speech are proposed. By using third order spectrum of speech signal to remove noise, in this algorithm the least spectrum difference to get refined pitch and the max harmonic number is given. And this algorithm utilizes spectral envelope to estimate signal-to-noise ratio of speech harmonics. Speech classification, voicing probability, and harmonic parameters of the voiced frame can be obtained. Simulation results indicate that the proposed algorithm, under complicated background noise, especially Gaussian noise, can effectively classify speech in high accuracy for voicing probability and the voiced parameters.     

Knowledge-based classification of polarimetric SAR images

In preparation for the flight of the Shuttle Imaging Radar-C (SIR-C) on board the Space Shuttle in the spring of 1994, a level-1 automatic classifier was developed on the basis of polarimetric SAR images acquired by the JPL AirSAR system. The classifier uses L- and C-Band polarimetric SAR measurements of the imaged scene to classify individual pixels into one of four categories: tall vegetation (trees), short vegetation, urban, or bare surface, with the last category encompassing water surfaces, bare soil surfaces, and concrete or asphalt-covered surfaces. The classifier design uses knowledge of the nature of radar backscattering from surfaces and volumes to construct appropriate discriminators in a sequential format. The classifier, which was developed using training areas in a test site in Northern Michigan, was tested against independent test areas in the same test site and in another site imaged three months earlier. Among all cases and all categories, the classification accuracy ranged between 91% and 100%     

Laboratory backscatter measurements over urea ice with a snow coverat Ku band

Indoor laboratory facilities were used to measure radar backscatter at Ku band (13.9 GHz) over urea ice, which has been shown to be structurally similar to sea ice. Data were collected at angles of incidence from normal to 55°, over very thin (0 to 9 cm) ice, snow-covered ice, and ice with a hooded snow cover. The laboratory proved to be useful in creating and controlling specific physical properties of ice while keeping all other variables constant, a difficulty with measurements collected in the field. It was found that surface scattering and the dielectric constant are the dominant factors that cause variations (up to 15 dB) in the measured backscatter. The addition of a snow cover increased the surface roughness of the smooth ice, increasing the backscatter at 20° incidence angle by about 11 dB and decreasing the backscatter at normal incidence by about 6 dB. The subsequent flooding of this snow layer increased the backscatter at all angles of incidence due to the increased dielectric constant of the wet slush layer. These results indicate the importance of the snow layer in influencing the surface characteristics of the ice sheet, which in turn modifies the backscattered signal