基于序列非线性滤波SAR影像水体自动提取

由于合成孔径雷达（SAR）图像中固有斑点噪声的强烈影响，某些对于光学图像有很好效果的目标提取区域分割方法，但是对于SAR图像来说，效果不很好，针对水体目标的亮度及形状分布特征，不作阈值分割处理，而采用序列非线性滤波处理方法，可以快速有效地实现SAR影像水体目标的自动提取识别。     

基于显著性分析的SAR溢油图像中暗斑区域检测方法研究

为了提高合成孔径雷达(syntheticapertureradar,SAR)图像中海上溢油的检测精度,设计了一种图像显著性和均值偏移分割相结合的暗斑检测方法,验证了该方法在SAR溢油图像暗斑识别中的有效性,为溢油区域的准确检测奠定了基础。针对SAR溢油图像中暗斑区域在人类视觉中属于感兴趣区域的特性,利用图像显著性检测方法对整幅SAR图像进行处理生成显著性图像,然后利用均值偏移方法在SAR图像的Lab空间将图像分割成不同区域,最后在显著性图像上利用每个区域的平均显著性值与动态阈值之间的关系确定溢油图像中的暗斑区域。实验结果表明,该方法不需要人工交互,且能够有效地提取出SAR图像溢油中的暗斑区域。     

海洋溢油合成孔径雷达图像特征提取及其关键度分析

就SAR图像溢油检测的方法论而言,用于识别溢油和疑似现象的定性或定量的统计特征量选择,通常是任意的。对于不同的分类模型,所选用的特征量也不尽相同。主要是进行海洋SAR图像特征提取及其关键度分析。其目的是将"最小距离"判别法应用于海上溢油和疑似溢油的识别研究。首先,针对海洋SAR图像溢油检测常用的特征量,进行冗余处理;然后,引入关键系数,定量地研究特征量的关键度,提取显著特征量;藉以构造一个多维的特征矢量空间,以适于最小距离判别法在特征矢量空间中进行溢油和疑似溢油的识别研究。     

基于改进Faster R-CNN模型的SAR图像溢油检测方法

SAR（synthetic aperture radar）图像溢油暗斑准确识别对海上溢油应急工作具有重要的意义。为减少SAR图像特征提取、特征选择过程中人为因素对溢油检测精度的影响,本文将Faster R-CNN卷积神经网络模型引入SAR图像溢油检测并进行了改进。针对溢油暗斑形状多样及SAR图像背景复杂的特点,选用结构一致且实用性强的VGG16卷积网络获取图像特征,并使用软化非极大值抑制算法（Soft-NMS）进行优化。同时基于相同的数据集,提取常用的SAR图像几何特征、灰度特征和纹理特征,构建反向传播（backpropagation,BP）人工神经网络溢油检测方法并与Faster R-CNN方法进行对比。实验结果表明,基于改进Faster-RCNN模型的溢油检测方法溢油检测率达到0.78,且溢油检测虚警率低于0.25,相比BP人工神经网络溢油检测方法样本识别率、溢油检测率分别提高了4%和5%,溢油虚警率降低了5%。     

基于DRLSE模型的SAR溢油提取方法

为提高海上溢油轮廓SAR提取精度,验证了FCM(Fuzzy C-Means Algorithm)与DRLSE(Distance Regularized Level Set Evolution)模型结合的方法提取SAR溢油信息的有效性;鉴于其无法避免细小噪音的影响以及薄油膜提取效果不好的问题,提出了阈值和DRLSE模型结合的溢油信息提取方法,通过阈值构建溢油区域初始轮廓,克服了图像细小噪声对溢油提取的影响,更有利于提取薄油膜信息,溢油提取精度优于H/A/alpha-Wishart非监督分类方法和FCM与DRLSE模型结合的方法。     

不同分辨率合成孔径雷达舰船检测中杂波模型适用性分析

基于恒虚警率的方法是经典的舰船检测算法,其核心是对海杂波统计分布的精确建模。随着合成孔径雷达(SAR)分辨率的提高,海杂波分布不再符合大数定理,而是表现为长拖尾形状,现已发展了诸如韦布尔分布、对数正态分布、G0分布、K-root分布、广义伽马分布等多种海杂波模型。本文首先利用上述模型对不同分辨率的SAR图像进行杂波建模,并在此基础上进行基于恒虚警率方法的舰船检测,然后,利用K-L距离和ROC曲线来评价各模型对不同分辨率条件下SAR海杂波拟合的精度与舰船检测的性能,给出了不同分辨率SAR适用于舰船检测的最优分布模型。     

基于深度学习的海浪SAR图像分类

针对海浪场中合成孔径雷达(SAR)图像的灰度特征混杂、人工目视分类困难的问题,本文利用简化的Inception-ResNet-V2模型与注意力机制相结合的方法,在减少网络层数、加快运算效率的同时,大幅度提升了计算机对SAR图像中海浪条纹清晰度的识别准确率。在利用模型进行图像分类时,本文提出分块识别的方式,对各块分类概率结果进行累加后取概率最大的类别,可提升1.8%的识别准确率,使最终准确率达到了89.6%。最后,本文基于深度学习结合12 000幅分类样本实现了总计5万幅ASAR图像的分类研究,获得条纹清晰组11 069幅、中间组16 560幅和不清晰组22 371幅分类图像。     

基于SAR极化比和纹理特征的海面溢油识别方法

针对海洋表面SAR影像的特点，采用基于灰度共生矩阵的纹理特征方法是提取海面溢油信息的常用方法，但实际海洋表面复杂的信息使得SAR图像上产生类似溢油现象的暗斑区域，这导致在利用纹理特征方法提取溢油信息时存在虚警率，降低了溢油信息的提取精度。基于RADARSAT-2 SAR四极化影像，本文提出基于SAR极化比影像的纹理特征识别方法对海面油膜进行识别提取。结果显示，基于SAR极化比影像的纹理特征识别方法可以有效且准确地提取海面溢油信息，相比于VV极化影像的纹理特征识别方法，溢油监测过程中的虚警率降低了17.96%，溢油监测总体精度达到96.83%。     

基于模型相似度拟合的海杂波统计方法

本文提出一种基于模型相似度拟合的海杂波统计方法。首先根据合成孔径雷达(SAR)图像计算瑞利分布、对数正态分布、韦布尔分布、K分布、G0分布5种经典的海杂波分布的概率密度函数,然后根据模型间的相似度准则拟合得到新的海杂波分布模型。文章利用四景不同类型的真实SAR数据对算法的拟合性能进行了评价,结果显示利用该算法得到的拟合模型与真实SAR数据的平均Kullback-Leibler距离仅为0.015 84,远优于其他分布模型。基于该拟合模型的恒虚警率舰船检测算法对四景SAR数据的平均检测精度达到95.75%,在控制虚警和漏检方面均优于采用其他模型的同类方法。     

合成孔径雷达图像中海洋内波的特征检测

海洋内波在合成孔径雷达(SAR)图像中具有特定纹理分布,因此可以根据海洋内波SAR图像的纹理分布特性,进行内波海面特征检测。利用功率谱对具有不同纹理特征的SAR图像进行分析,并对特征进行分类统计得出内波的功率谱特征。在海洋内波区域,利用多孔小波变换及多尺度融合提取海洋内波特征,在此基础上对波-流散射系数进行提取。提出了"内波特征的功率谱特征-多孔小波综合检测算法",为海洋内波参数反演奠定了基础。     

Extremely deep-draft iceberg scouring in the glacial North Atlantic Ocean

Side-scan sonar and sub-bottom profiling investigations of the Iceland-Faeroe Ridge have revealed the widespread presence of iceberg plowmarks at water depths between 750 and 860 m. This is well beyond the maximum water depth reported for ice scouring in the adjacent areas of the North Atlantic. An exceptionally deep iceberg plowmark was found at the southern flank of the ridge, which could be traced down to 940 m water depth. Supported by published nearby sediment core data, a Saalian (MIS 6) age of that iceberg plowmark is suggested. Considering a glacial sea level of 120 m below present, the paleo-draft of this iceberg is estimated to have been at least 820 m. The keel depth of such giant icebergs drifting in the glacial northern North Atlantic significantly exceeds the draft dimensions of large icebergs seen today in the Antarctic. The most likely source area for this extreme deep-draft type of iceberg is inferred to be southern Greenland. Plowmark direction, which in our case indicates iceberg entrainment by the North Atlantic Current, is suggested to be a useful indicator of (sub)surface paleo-circulation during late Pleistocene iceberg surging and deglaciation episodes.     

Iceberg drifting and distribution in the Vilkitsky Strait studied by detailed satellite radar and optical images

This paper is devoted to the detection and identification of icebergs in the Russian Arctic Seas from the use of high- and medium-resolution radar and optical images from EROS-B, Radarsat-1, Radarsat-2, SPOT-4 and SPOT-5 Earth observation satellites. In July–September of 2011–2013, the SCANEX Research and Development Center, the Federal State Unitary Enterprise Atomflot, and other partner organizations provided operational satellite monitoring of icebergs in the Kara Sea and the Laptev Sea. More than 130 highly detailed optical and radar images were received and processed. The Vilkitsky Strait—one of the narrowest and most dangerous places within the Northern Sea Route—was chosen as an experimental polygon. As a result, iceberg location in the strait during the 2011–2013 navigation periods was analyzed, as were the iceberg size, area, drift direction, and height.     

基于分区自适应阈值的声呐障碍物检测算法研究

针对UUV避碰声呐探测障碍物过程中自主选择分割阈值进行障碍物检测的问题,提出了基于分区自适应阈值的障碍物检测算法。首先将避碰声呐图像均匀分为相同大小的图像块,对每个图像块基于最大类间方差算法确定该区域障碍物图像分割的高低阈值,然后对检测到的障碍物进行形态学处理去除孤立噪声点,对目标区域进行连通性分析及内部空洞处理,最终得到完整的障碍物轮廓信息。通过湖试数据验证表明了该方法对声呐障碍物检测的有效性。     

The effect of dynamic–thermodynamic icebergs on the Southern Ocean climate in a three-dimensional model

Melting icebergs are a mobile source of fresh water as well as a sink of latent heat. In most global climate models, the spatio-temporal redistribution of fresh water and latent heat fluxes related to icebergs is parameterized by an instantaneous more or less arbitrary flux distribution over some parts of the oceans. It is uncertain if such a parameterization provides a realistic representation of the role of icebergs in the coupled climate system. However, icebergs could have a significant climate role, in particular during past abrupt climate change events which have been associated with armada’s of icebergs. We therefore present the interactive coupling of a global climate model to a dynamic thermodynamic iceberg model, leading to a more plausible spatio-temporal redistribution of fresh water and heat fluxes. We show first that our model is able to reproduce a reasonable iceberg distribution in both hemispheres when compared to recent data. Second, in a series of sensitivity experiments we explore cooling and freshening effects of dynamical icebergs on the upper Southern Ocean and we compare these dynamic iceberg results to the effects of an equivalent parameterized iceberg flux.In our model without interactive icebergs, the parameterized fluxes are distributed homogeneously South of 55°S, whereas dynamic icebergs are found to be concentrated closer to shore except for a plume of icebergs floating North–East from the tip of the Antarctic Peninsula. Compared to homogeneous fluxes, the dynamic icebergs lead to a 10% greater net production of Antarctic bottom water (AABW). This increased bottom water production involves open ocean convection, which is enhanced by a less efficient stratification of the ocean when comparing to a homogeneous flux distribution.Icebergs facilitate the formation of sea-ice. In the sensitivity experiments, both the fresh water and the cooling flux lead to a significant increase in sea-ice area of 12% and 6%, respectively, directly affecting the highly coupled and interactive air/sea/ice system. The consequences are most pronounced along the sea-ice edge, where this sea-ice facilitation has the greatest potential to affect ocean stratification, for example by heat insulation and wind shielding, which further amplifies the cooling and freshening of the surface waters.     

Surface buoyant plumes from melting icebergs in the Labrador Sea

Canada׳s Department of Fisheries and Oceans (DFO) conducts annual surveys in the Labrador Sea along the repeat hydrography line AR7W. The occupation of the AR7W line in May 2013 was followed by the experiment aimed at resolving the imprint of melting drifting icebergs on the upper layer thermohaline characteristics in the Labrador Sea. We present high-resolution observations around two icebergs conducted with the towed undulating platform Moving Vessel Profiler (MVP). The first iceberg drifted in relatively warm water of Atlantic origin (~2.5–3.1 °C) off Greenland, while the second iceberg was on the Labrador shelf in cold water below 0 °C. Both icebergs had a lengthscale of O(100 m). In both cases surface buoyant plumes fed by melt water and attached to the iceberg were observed. The plumes were evident in the anomalous thermohaline characteristics of the seawater. Their density anomalies were sufficiently strong to produce visible frontal structures, which imply a development of the intrinsic dynamics associated with a plume. The first plume formed over a time interval of ~10 h, while the second plume formed over several days and extended for more than 1 km (tenfold the iceberg׳s size). Strong vertical displacements of the pycnocline were observed near the second iceberg. They are interpreted as the internal wave wake. This interpretation is based on the temporal scale of these oscillations (local buoyancy frequency), as well as on the spatial orientation of these waves with respect to the iceberg drift relative to the pycnocline. The observed internal waves partially overlapped with the plume and affected its structure. The saline seawater splashing by swell contributed to the surface melting of the icebergs. Scaling analysis of the second plume suggests that it could be in the “rotational” dynamic regime with recirculating anticyclonic flow.     

Real-time radar detection of iceberg shadows

The Hough transform is used to estimate the radial direction of shadows cast by icebergs in surface-based marine radar returns using a Warp systolic computer. Automatically locating these shadows leads directly to automatically locating the icebergs casting the shadows. This is useful for the unattended navigation of a vessel through sea ice, as well as for improving the accuracy of a classified sea-ice image. The results reported show that the Hough transform can successfully estimate the radial directions of the shadows cast by the icebergs, and that the implementation on the Warp computer is commensurate with a real-time requirement for shadow detection     

Impact of drifting icebergs on surface phytoplankton biomass in the Southern Ocean: Ocean colour remote sensing and in situ iceberg tracking

Icebergs that calve from the Antarctic ice shelves and drift in the Southern Ocean melt to deliver fresh water, dust and minerogenic particles to the surface ocean along the iceberg's path. Each of these components may have an effect on growth conditions for phytoplankton, as might the mechanical effects of the iceberg keel disturbing the water. Although anecdotal and small-scale surveys suggest that drifting icebergs increase local primary production, no large-scale studies have been reported. An analysis of satellite and automated iceberg tracking data from the Weddell Sea, covering the months October to March, from 1999 to 2004, showed that the probability of increased surface phytoplankton biomass was up to one-third higher in the wake of a tracked iceberg compared to background biomass fluctuations. Only during the month of February were the effects of icebergs on surface biomass likely to be negative, whereas background biomass fluctuations were likely to be negative during March. These results confirm icebergs as a factor affecting phytoplankton in the Southern Ocean and highlight the need for detailed process studies so that responses to future changes in the Antarctic ice sheets may be predicted.     

Remote sensing of icebergs by ground-wave Doppler radar

A software detection model has been developed to predict the returned Doppler spectrum for an iceberg target for ground-wave Doppler radars. This software model is based on proposed new estimates for the backscattered Doppler-dependent iceberg cross section for assumed iceberg models, as well as the backscattered Doppler spectrum from the ocean surface. The model includes estimates for forward and reverse transmission losses, based on classical spherical earth derivations. In addition, the transmission losses account for the effects of surface roughness through a modified surface impedance. Standard estimates for man-made and atmospheric noise have been considered in the detection model. A comparison between the results predicted by the detection model and data acquired during an experiment conducted at Byron Bay, Labrador, Canada has been effected. The hardware used for the experiment was an HF Doppler radar operating at 25.40 MHz. The transmitting antenna was a three-element Yagi array and the receiving antenna a 24-element narrow-beam linear array. By using iceberg ground truthing information the Doppler spectrum for individual icebergs was predicted using the software model. The predicted spectra were compared with the received spectra on a target signal-to-noise power ratio basis. The results of this comparison give a degree of confidence to the detection model and show that ground-wave radars are effective ice hazard remote sensors.     

一种快速非局部平均合成孔径雷达海洋图像斑点噪声抑制方法

合成孔径雷达在海洋环境监测和海洋研究中扮演着越来越重要的角色。受其成像机制的影响，合成孔径雷达图像总是受到斑点噪声的污染。斑点噪声的存在会增大目标识别、跟踪和分类的难度，也会降低雷达信号的信噪比。合成孔径雷达海洋图像具有一些特殊的性质：海洋现象在雷达图像中主要呈现为条带状或斑块状的结构。这些条带状或斑块状的结构呈现出高度的自相似性或信息冗余。非局部平均方法能够衡量图像中不同图像块之间纹理结构的相似性，并利用图像的自相似性对图像进行去噪。但非局部平均去燥方法存在计算量巨大、计算耗时长的缺点，这几乎限制了其实际应用。本文采用一种自适应方法将雷达图像中的像素点区分为纹理区像素点和平坦区像素点。对纹理区像素点，采用较大的相似窗和搜索窗，对平坦区像素点，采用较小的相似窗和搜索窗，从而提高计算速度。进一步，本文基于计算统一设备并行架构（CUDA）技术，利用计算机图形处理器（GPU）对前述算法进行并行加速。与经典非局部平均算法相比，加速后算法的计算效率提高了200倍。     

绿潮Lansat影像滑动窗口自适应阈值全自动检测方法

Since the atmospheric correction is a necessary preprocessing step of remote sensing image before detecting green tide, the introduced error directly affects the detection precision. Therefore, the detection method of green tide is presented from Landsat TM/ETM plus image which needs not the atmospheric correction. In order to achieve an automatic detection of green tide, a linear relationship(y =0.723 x+0.504) between detection threshold y and subtraction x(x=λ_(nir)–λ_(red)) is found from the comparing Landsat TM/ETM plus image with the field surveys.Using this relationship, green tide patches can be detected automatically from Landsat TM/ETM plus image.Considering there is brightness difference between different regions in an image, the image will be divided into a plurality of windows(sub-images) with a same size firstly, and then each window will be detected using an adaptive detection threshold determined according to the discovered linear relationship. It is found that big errors will appear in some windows, such as those covered by clouds seriously. To solve this problem, the moving step k of windows is proposed to be less than the window width n. Using this mechanism, most pixels will be detected[n/k]×[n/k] times except the boundary pixels, then every pixel will be assigned the final class(green tide or sea water) according to majority rule voting strategy. It can be seen from the experiments, the proposed detection method using multi-windows and their adaptive thresholds can detect green tide from Landsat TM/ETM plus image automatically. Meanwhile, it avoids the reliance on the accurate atmospheric correction.