The polarimetric features of oil spills in full polarimetric synthetic aperture radar images

Compared with single-polarized synthetic aperture radar （SAR） images, full polarimetric SAIl images contain not only geometrical and backward scattering characteristics, but also the polarization features of the scattering targets. Therefore, the polarimetric SAR has more advantages for oil spill detection on the sea surface. As a crucial step in the oil spill detection, a feature extraction directly influences the accuracy of oil spill discrimination. The polarimetric features of sea oil spills, such as polarimetric entropy, average scatter angle, in the full polarimetric SAR images are analyzed firstly. And a new polarimetric parameter P which reflects the proportion between Bragg and specular scattering signals is proposed. In order to investigate the capability of the polarimetric features for observing an oil spill, systematic comparisons and analyses of the multipolarization features are provided on the basis of the full polarimetric SAR images acquired by SIR-C/X-SAR and Radarsat-2. The experiment results show that in C-band SAR images the oil spills can be detected more easily than in L-band SAR images under low to moderate wind speed conditions. Moreover, it also finds that the new polarimetric parameter is sensitive to the sea surface scattering mechanisms. And the experiment results demonstrate that the new polarimetric parameter and pedestal height perform better than other polarimetric parameters for the oil spill detection in the C-band SAR images.

Unsupervised change detection of man-made objects using coherent and incoherent features of multi-temporal SAR images

Constrained by complex imaging mechanism and extraordinary visual appearance,change detection with synthetic aperture radar(SAR)images has been a difficult research topic,especially in urban areas.Although existing studies have extended from bi-temporal data pair to multi-temporal datasets to derive more plentiful information,there are still two problems to be solved in practical applications.First,change indicators constructed from incoherent feature only cannot characterize the change objects accurately.Second,the results of pixel-level methods are usually presented in the form of the noisy binary map,making the spatial change not intuitive and the temporal change of a single pixel meaningless.In this study,we propose an unsupervised man-made objects change detection framework using both coherent and incoherent features derived from multi-temporal SAR images.The coefficients of variation in timeseries incoherent features and the man-made object index(MOI)defined with coherent features are first combined to identify the initial change pixels.Afterwards,an improved spatiotemporal clustering algorithm is developed based on density-based spatial clustering of applications with noise(DBSCAN)and dynamic time warping(DTW),which can transform the initial results into noiseless object-level patches,and take the cluster center as a representative of the man-made object to determine the change pattern of each patch.An experiment with a stack of 10 TerraSAR-X images in Stripmap mode demonstrated that this method is effective in urban scenes and has the potential applicability to wide area change detection.

Coherent change detection of fine traces based on multi-angle SAR observations

Coherent change detection(CCD) is an effective method to detect subtle scene changes that occur between temporal synthetic aperture radar(SAR) observations. Most coherence estimators are obtained from a Hermitian product based on local statistics. Increasing the number of samples in the local window can improve the estimation bias, but cause the loss of the estimated images spatial resolution. The limitations of these estimators lead to unclear contour of the disturbed region, and even the omission of fine change targets. In this paper, a CCD approach is proposed to detect fine scene changes from multi-temporal and multi-angle SAR image pairs. Multi-angle CCD estimator can improve the contrast between the change target and the background clutter by jointly accumulating singleangle alternative estimator results without further loss of image resolution. The sensitivity of detection performance to image quantity and angle interval is analyzed. Theoretical analysis and experimental results verify the performance of the proposed algorithm.

A Hybrid Features Based Detection Method for Inshore Ship Targets in SAR Imagery

Convolutional Neural Networks(CNNs)have recently attracted much attention in the ship detection from Synthetic Aperture Radar(SAR)images.However,compared with optical images,SAR ones are hard to understand.Moreover,due to the high similarity between the man-made targets near shore and inshore ships,the classical methods are unable to achieve effective detection of inshore ships.To mitigate the influence of onshore ship-like objects,this paper proposes an inshore ship detection method in SAR images by using hybrid features.Firstly,the sea-land segmentation is applied in the pre-processing to exclude obvious land regions from SAR images.Then,a CNN model is designed to extract deep features for identifying potential ship targets in both inshore and offshore water.On this basis,the high-energy point number of amplitude spectrum is further introduced as an important and delicate feature to suppress false alarms left.Finally,to verify the effectiveness of the proposed method,numerical and comparative studies are carried out in experiments on Sentinel-1 SAR images.

Detecting spatio-temporal urban surface changes using identified temporary coherent scatterers

Synthetic aperture radar(SAR) is able to detect surface changes in urban areas with a short revisit time, showing its capability in disaster assessment and urbanization monitoring.Most presented change detection methods are conducted using couples of SAR amplitude images. However, a prior date of surface change is required to select a feasible image pair. We propose an automatic spatio-temporal change detection method by identifying the temporary coherent scatterers. Based on amplitude time series, χ^(2)-test and iterative single pixel change detection are proposed to identify all step-times: the moments of the surface change. Then the parameters, e.g., deformation velocity and relative height, are estimated and corresponding coherent periods are identified by using interferometric phase time series. With identified temporary coherent scatterers, different types of temporal surface changes can be classified using the location of the coherent periods and spatial significant changes are identified combining point density and F values. The main advantage of our method is automatically detecting spatio-temporal surface changes without prior information. Experimental results by the proposed method show that both appearing and disappearing buildings with their step-times are successfully identified and results by ascending and descending SAR images show a good agreement.

STUDY ON THE TECHNIQUE TO DETECT TEXTURE FEATURES IN SAR IMAGES

This letter studies on the detection of texture features in Synthetic Aperture Radar (SAR) images. Through analyzing the feature detection method proposed by Lopes, an improved texture detection method is proposed, which can not only detect the edge and lines but also avoid stretching edge and suppressing lines of the former algorithm. Experimental results with both simulated and real SAR images verify the advantage and practicability of the improved method.

A NOVEL SHIP WAKE DETECTION METHOD OF SAR IMAGES BASED ON FREQUENCY DOMAIN

Moving ships produce a set of waves of "V' pattern on the ocean. These waves can often be seen by Synthetic Aperture Radar (SAR). The detection of these wakes can provide important information for surveillance of shipping, such as ship traveling direction and speed. A novel approach to the detection of ship wakes in SAR images based on frequency domain is provided in this letter. Compared with traditional Radon-based approaches, computation is reduced by 20%-40% without losing nearly any of detection performance. The testing results using real data and simulation of synthetic SAR images test the algorithm's feasibility and robustness.

基于局部显著特征聚焦学习的SAR舰船智能检测

合成孔径雷达(Synthetic Aperture Radar,SAR)图像的船舶目标检测,因其广泛的应用前景而备受关注。近年来,基于深度学习的SAR图像船舶目标检测在多种场景中表现出较好性能。然而,由于SAR独特的成像机制,舰船目标通常与背景环境具有相似的散射特性使得实际的船舶目标难以辨识,且船舶目标尺度较小,导致准确检测船舶目标具有挑战性。为了缓解这一问题,本文提出了一种基于局部显著特征聚焦学习的SAR舰船检测方法。首先,设计了双重注意力模块,通过对通道级和空间级的特征进行双重注意力加权,以充分地探索船舰目标的关键语义特征,从而提升模型的深度提取能力。随后,为了进一步提升模型对船舶目标特征的表征能力,设计了平衡特征金字塔网络模块,通过对舰船目标的多尺度特征进行缩放、增强和聚合处理,以实现多尺度特征间的语义和空间信息均衡分布。最后,在SAR舰船检测数据集(SAR Ship Detection Dataset,SSDD)上进行了广泛的实验分析,实验结果一致性地证明了所提方法在提升SAR图像舰船目标检测准确性方面的有效性。

结合特征融合和注意力机制的SAR舰船检测算法

针对现有的合成孔径雷达目标检测算法仅利用图像底层特征进行检测存在的对小尺度舰船目标的检测率较低问题,提出一种结合特征融合和注意力机制的目标检测算法。面向SAR舰船目标检测,在原始主干网络SSD目标检测算法的基础上,引入注意力机制模块、不同层次的特征图进行特征融合、对含有小尺度目标的图像进行过采样还通过多次复制粘贴小目标实现数据增广。实验通过对SAR舰船图像数据集的大量训练和测试,结果表明本文算法能有效提升对舰船目标的综合检测性能,在公开SAR舰船目标检测数据集上平均精度可以达到94.16%。

语义增强与高阶强交互的SAR图像舰船检测

合成孔径雷达(synthetic aperture radar,SAR)图像背景信息复杂、舰船目标边缘模糊,且多为容易丢失的小尺度舰船目标。针对上述问题,提出语义增强与高阶强交互的SAR图像舰船检测。该方法利用部分卷积与非对称卷积构建部分非对称卷积聚合网络,在减少计算复杂度、轻量化主干网络的同时,更好地捕捉多尺度舰船特征,同时在上采样部分引入双层路由注意力,增强对图像上下文信息的利用。另外,通过递归的方式进行特征提取,可以较好解决区域内信息交互的问题,实现不同级别特征之间的高阶交互建模,提升模型检测能力。在公开的HRSID遥感数据集上进行实验的结果表明,该方法的检测精度达到91.23%,相比原模型提升5.13%,准确率与召回率分别提升2.41%和7.16%,与主流算法相比具有较好的检测效果。

SAR图像舰船目标检测的轻量化和特征增强研究

针对合成孔径雷达(SAR)图像中的舰船目标的准确率易受近岸杂波的影响,且现有检测算法复杂度高,在嵌入式设备上的部署难度大的问题,提出一种适用于嵌入式设备的轻量化高精度SAR图像舰船目标检测算法CA-Shuffle-YOLO。基于YOLO v5目标检测算法,对骨干网络进行轻量化及特征精细化提取两个方面的改进,引入轻量化模块以降低网络的计算复杂度,提高推理速度,并引入协同注意力机制模块增强算法对近岸船舶目标的细节信息的提取能力。在特征融合网络中采用加权特征融合以及跨模块融合,增强模型对SAR舰船目标的细节信息的融合能力,同时,利用深度卷积模块降低计算复杂度,提高实时性。通过在SSDD舰船目标检测数据集上的测试及对比实验的结果,表明CA-Shuffle-YOLO的检测准确率约为97.4%,检测帧率为206 FPS,所需运算复杂度为6.1 GFlops,相比原始的YOLO v5,所提方法的检测帧率提升了60 FPS,所需运算复杂度降低为原来的12%。

基于长短路融合及数据平衡的SAR船舶检测算法

针对SAR图像检测船舶任务中的目标小、近岸样本目标检测困难等问题,文章提出一种名为长短路特征融合网络(Long and Short path Feature Fusion Network,LSFF-Net)的船舶检测网络。该网络通过长短路特征融合模块有效协调了大目标与小目标检测,避免小目标特征信息的丢失。网络中应用结构重参数化结构提高了模块学习能力。为了满足多尺度目标检测,加入特征金字塔网络,融合多尺度特征。为了应对近岸样本目标检测,设计数据重分配算法,提高了对近岸样本目标的检测精度。实验结果表明:在公开数据集检测时,算法的平均精度(Average Precision,AP)达到97.50%,优于主流目标检测算法。该方法为提高SAR图像中小目标和近岸样本目标检测精度提供了新的实现方案。

基于注意力和自适应特征融合的SAR图像飞机目标检测

针对合成孔径雷达(Synthetic Aperture Radar,SAR)图像中飞机目标尺度多样性及背景强散射干扰的问题,提出了一种基于坐标注意力和自适应特征融合的YOLOv4 SAR图像飞机目标检测算法。该方法首先在主干网络引入坐标注意力机制,以增强对于飞机散射点组合结构的聚焦能力以及抗背景干扰能力。其次,在特征增强网络中引入自适应特征融合机制,提高了对不同大小飞机的特征提取能力,同时改善了YOLOv4算法召回率和精确率不平衡的问题。最后,通过改进的K-Means聚类针对飞机目标调整先验框的尺寸,提高了模型的定位精度。实验结果表明,改进算法召回率达到91.01%,精确率达到90.09%,AP 0.5达到92.34%,分别较原YOLOv4算法提高2.49%,6.56%和3.62%。

基于全局位置信息和残差特征融合的SAR船舶检测算法

针对合成孔径雷达(synthetic aperture radar,SAR)图像船舶目标尺度不一且易受海面、地面杂波和相干斑噪声的影响,难以提取目标多维特征且特征融合过程中易产生语义歧义,造成船舶目标检测率低,虚警率高的问题,提出一个基于全局位置信息和残差特征融合的SAR船舶目标检测算法。基于Faster区域卷积神经网络(region convolutional neural network,R-CNN)目标检测算法,在特征提取网络和特征融合网络中进行改进:在特征提取网络中使用高宽注意力机制提取目标在图像中的全局位置信息,增强目标的多维特征提取能力;在特征融合网络中使用带有残差连接的双向特征金字塔网络削弱特征融合过程中的语义歧义,降低复杂背景下的船舶目标虚警率,同时进行不同层级的多尺度特征双向融合,增强高低层特征的联系,提升多尺度船舶目标的检测能力。在SAR船舶数据集上达到98.2%的均值平均精度,超过部分算法2.4%以上。实验表明,所提算法有效提取了目标的多维特征,显著缓解了语义歧义问题,具有较好的检测能力和泛化能力。

基于特征解耦的SAR图像舰船检测蒸馏

目前,基于深度学习的合成孔径雷达(SAR)舰船目标检测方法受到广泛关注。但因为模型参数量大、运算内存高等问题限制了其实际应用。通过学生网络模仿教师网络,知识蒸馏被视作一种高效的模型压缩方法。然而,大部分的知识蒸馏算法只针对常见的可见光图像任务,将其直接应用到复杂的SAR图像舰船目标检测上性能表现不佳。通过分析,出现上述性能不佳现象有以下两个原因:(1)前景背景面积严重失衡;(2)缺乏对前景和背景像素的关系建模。针对上述问题,提出基于解耦特征的拓扑距离知识蒸馏算法。前景和背景解耦蒸馏可以缓解前景背景失衡问题。通过解耦特征拓扑距离蒸馏,学生网络可以从教师网络学习到前景背景之间的关系,增强对背景噪声鲁棒性。实验结果表明,相比许多蒸馏算法,所提出的算法可以十分有效地提升学生网络在SAR图像舰船目标检测精度。比如,基于ResNet18-C4骨干网络的Faster R-CNN模型在HRSID数据集上AP提升6.85个百分点,从31.81%提升到38.66%。

共性显著特征引导的SAR图像交通感兴趣区域检测

由于独特的成像机制,SAR图像目标检测存在缺乏光谱信息、斑点噪声干扰和方位角敏感等问题。在光学图像中使用的显著性目标检测方法,大多数不适合直接应用在SAR图像的目标检测中。针对上述问题,从视觉显著性角度入手,提出一种针对SAR图像的联合多图共性显著性分析和单图显著性分析的交通感兴趣区域检测算法。首先提出一种增强的方向平滑滤波器,有效降低SAR图像的斑点噪声。然后,针对多幅图像,提出一个共性显著分析模型,通过提取曲线、纹理和亮度特征等共性特征再聚类,得到共性显著图;针对单幅图像,通过共生直方图生成单图显著图。最后通过一种新颖的融合方法实现最终的多幅图像显著性检测。试验在一个数据集上与4种显著性分析模型进行了比较,模型的ROC曲线位于最上方,曲线下面积为0.909 5,综合评价指标F-Measure值为0.674 45,能准确识别显著区域,抑制背景信息,不会产生误判;与3种经典有效的SAR图像目标检测方法进行了比较,此算法在获得完整和准确的感兴趣区域描述方面表现出色,在比较方法中表现出最高的定位精度。显著性目标检测模型有效利用SAR图像的特征信息并抑制干扰信息,使SAR图像交通感兴趣区域检测和车辆检测结果更准确。

基于改进YOLOv5的船舶多尺度SAR图像检测算法

针对复杂场景下合成孔径雷达(Synthetic aperture radar, SAR)图像船舶目标像素尺度差异大和船舶密集排列造成目标漏检的问题,提出一种基于改进YOLOv5的船舶多尺度SAR图像检测算法。对于YOLOv5的颈部网络,采用双向特征金字塔结构(Bi-directional feature pyramid network, BiFPN)提升网络多尺度特征融合能力,并在其自下而上的特征融合支路中,基于深度可分离卷积(Depthwise separable convolution, DSC)和通道MLP构建EC-MLP(Enhanced channel-MLP)模块,从而丰富语义信息,提供更充分的船舶目标上下文特征;引入全局注意力机制(Global attention mechanism, GAM),使网络对输入特征进行针对性提取并运算,减少网络的信息丢失;此外,使用SIoU损失函数进一步提高网络的训练收敛速度和检测精度。在SSDD和HRSID数据集上与其他8种方法(Faster R-CNN、Libra R-CNN、FCOS、YOLOv5s、PP-YOLOv2、YOLOX-s、PP-YOLOE-s和YOLOv7-tiny)进行对比实验。实验结果表明:改进后算法在SSDD数据集上的AP50达到了96.7%,在HRSID数据集上AP50达到了95.6%,优于对比方法。

高分三号卫星影像在地表变化检测中的应用试验

合成孔径雷达(SAR)影像受相干斑噪声、成像几何畸变和背景信息复杂等因素影响,在地表变化检测领域尚未得到广泛应用。文章研究总结了目前常用的基于SAR影像的变化检测方法及技术流程,构建“高分三号”(GF-3)卫星SAR影像差分干涉测量和极化数据变化检测技术方法与流程,并选择内蒙古新井露天煤矿区域进行试验,成功验证了GF-3影像应用于地表变化检测的技术可行性和结果可靠性。主要结论如下:1)基于GF-3卫星影像差分干涉测量成功获取新井煤矿范围形变图,形变较大区域与光学影像解译结果基本吻合,证明GF-3卫星具有差分干涉测量能力;2)基于GF-3影像变化检测提取的新井煤矿事故发生位置准确,提取变化区域面积为0.128 km^(2),与公开发布的监测结果(0.1 km^(2))基本一致,证明GF-3影像可应用于地表变化检测。

Field testing innovative differential geospatial and photogrammetric monitoring technologiesin mountainous terrain near Ashcroft,British Columbia,Canada

This paper presents a novel approach to continuously monitor very slow-moving translational landslides in mountainous terrain using conventional and experimental differential global navigation satellite system(d-GNSS)technologies.A key research question addressed is whether displacement trends captured by a radio-frequency“mobile”d-GNSS network compare with the spatial and temporal patterns in activity indicated by satellite interferometric synthetic aperture radar(InSAR)and unmanned aerial vehicle(UAV)photogrammetry.Field testing undertaken at Ripley Landslide,near Ashcroft in south-central British Columbia,Canada,demonstrates the applicability of new geospatial technologies to monitoring ground control points(GCPs)and railway infrastructure on a landslide with small and slow annual displacements(<10 cm/yr).Each technique records increased landslide activity and ground displacement in late winter and early spring.During this interval,river and groundwater levels are at their lowest levels,while ground saturation rapidly increases in response to the thawing of surficial earth materials,and the infiltration of snowmelt and runoff occurs by way of deep-penetrating tension cracks at the head scarp and across the main slide body.Research over the last decade provides vital information for government agencies,national railway companies,and other stakeholders to understand geohazard risk,predict landslide movement,improve the safety,security,and resilience of Canada’s transportation infrastructure;and reduce risks to the economy,environment,natural resources,and public safety.

利用GLCM纹理分析的高分辨率SAR图像建筑区检测

根据高分辨率SAR图像上建筑区的影像特征,提出了基于灰度共生矩阵(gray-level cooccurrence Matrix,GLCM)纹理分析的建筑区提取方法,该方法由初步定位和边界调整2个步骤组成,均遵循特征计算、基于Bhattacharyya距离的特征选择和KNN分类流程,所不同的是2个步骤中分别采用了逐块和逐点计算纹理特征的方式以兼顾纹理分析的效率和准确性。文中对不同SAR传感器获取的图像进行了实验。实验结果表明,选用具有最大Bhattacharyya距离值的3或4个特征可以获得较好的初步定位结果,建筑区的检测率超过80%,虚警率低于10%;随着边界调整的进行,检测到的建筑区边界逐渐接近于真实边界。实验结果验证了该算法的有效性。