多源特征自适应融合网络的高分遥感影像语义分割

目的 在高分辨率遥感影像语义分割任务中,仅利用可见光图像很难区分光谱特征相似的区域(如草坪和树、道路和建筑物),高程信息的引入可以显著改善分类结果。然而,可见光图像与高程数据的特征分布差异较大,简单的级联或相加的融合方式不能有效处理两种模态融合时的噪声,使得融合效果不佳。因此如何有效地融合多模态特征成为遥感语义分割的关键问题。针对这一问题,本文提出了一个多源特征自适应融合模型。方法 通过像素的目标类别以及上下文信息动态融合模态特征,减弱融合噪声影响,有效利用多模态数据的互补信息。该模型主要包含3个部分：双编码器负责提取光谱和高程模态的特征;模态自适应融合模块协同处理多模态特征,依据像素的目标类别以及上下文信息动态地利用高程信息强化光谱特征,使得网络可以针对特定的对象类别或者特定的空间位置来选择特定模态网络的特征信息;全局上下文聚合模块,从空间和通道角度进行全局上下文建模以获得更丰富的特征表示。结果 对实验结果进行定性、定量相结合的评价。定性结果中,本文算法获取的分割结果更加精细化。定量结果中,在ISPRS(International Society for Photogrammetry...     

面向遥感影像的深度语义哈希检索

目的 哈希检索旨在将海量数据空间中的高维数据映射为紧凑的二进制哈希码,并通过位运算和异或运算快速计算任意两个二进制哈希码之间的汉明距离,从而能够在保持相似性的条件下,有效实现对大数据保持相似性的检索。但是,遥感影像数据除了具有影像特征之外,还具有丰富的语义信息,传统哈希提取影像特征并生成哈希码的方法不能有效利用遥感影像包含的语义信息,从而限制了遥感影像检索的精度。针对遥感影像中的语义信息,提出了一种基于深度语义哈希的遥感影像检索方法。方法 首先在具有多语义标签的遥感影像数据训练集的基础上,利用两个不同配置参数的深度卷积网络分别提取遥感影像的影像特征和语义特征,然后利用后向传播算法针对提取的两类特征学习出深度网络中的各项参数并生成遥感影像的二进制哈希码。生成的二进制哈希码之间能够有效保持原始高维遥感影像的相似性。结果 在高分二号与谷歌地球遥感影像数据集、CIFAR-10数据集及FLICKR-25K数据集上进行实验,并与多种方法进行比较和分析。当编码位数为64时,相对于DPSH（deep supervised Hashing with pairwise labels）方法,在高分二号与谷歌地球遥感影像数据集、CIFAR-10数据集、FLICKR-25K数据集上,mAP（mean average precision）指标分别提高了约2%、6%7%、0.6%。结论 本文提出的端对端的深度学习框架,对于带有一个或多个语义标签的遥感影像,能够利用语义特征有效提高对数据集的检索性能。     

Wavelet-based image registration technique for high-resolution remote sensing images

Image registration is the process of geometrically aligning one image to another image of the same scene taken from different viewpoints at different times or by different sensors. It is an important image processing procedure in remote sensing and has been studied by remote sensing image processing professionals for several decades. Nevertheless, it is still difficult to find an accurate, robust, and automatic image registration method, and most existing image registration methods are designed for a particular application. High-resolution remote sensing images have made it more convenient for professionals to study the Earth; however, they also create new challenges when traditional processing methods are used. In terms of image registration, a number of problems exist in the registration of high-resolution images: (1) the increased relief displacements, introduced by increasing the spatial resolution and lowering the altitude of the sensors, cause obvious geometric distortion in local areas where elevation variation exists; (2) precisely locating control points in high-resolution images is not as simple as in moderate-resolution images; (3) a large number of control points are required for a precise registration, which is a tedious and time-consuming process; and (4) high data volume often affects the processing speed in the image registration. Thus, the demand for an image registration approach that can reduce the above problems is growing. This study proposes a new image registration technique, which is based on the combination of feature-based matching (FBM) and area-based matching (ABM). A wavelet-based feature extraction technique and a normalized cross-correlation matching and relaxation-based image matching techniques are employed in this new method. Two pairs of data sets, one pair of IKONOS panchromatic images from different times and the other pair of images consisting of an IKONOS panchromatic image and a QuickBird multispectral image, are used to evaluate the proposed image registration algorithm. The experimental results show that the proposed algorithm can select sufficient control points semi-automatically to reduce the local distortions caused by local height variation, resulting in improved image registration results.     

Object detection in remote sensing images based on deep transfer learning

Multimedia Tools and Applications - Object detection is a basic part in remote sensing image processing. At present, it is more common to conduct the topic based on deep learning, however the...     

航空遥感图像深度学习目标检测技术研究进展

航空遥感图像目标检测旨在定位和识别遥感图像中感兴趣的目标,是航空遥感图像智能解译的关键技术,在情报侦察、灾害救援和资源勘探等领域具有重要应用价值。然而由于航空遥感图像具有尺寸大、目标小且密集、目标呈任意角度分布、目标易被遮挡、目标类别不均衡以及背景复杂等诸多特点,航空遥感图像目标检测目前仍然是极具挑战的任务。基于深度卷积神经网络的航空遥感图像目标检测方法因具有精度高、处理速度快等优点,受到了越来越多的关注。为推进基于深度学习的航空遥感图像目标检测技术的发展,本文对当前主流遥感图像目标检测方法,特别是2020—2022年提出的检测方法,进行了系统梳理和总结。首先梳理了基于深度学习目标检测方法的研究发展演化过程,然后对基于卷积神经网络和基于Transformer目标检测方法中的代表性算法进行分析总结,再后针对不同遥感图象应用场景的改进方法思路进行归纳,分析了典型算法的思路和特点,介绍了现有的公开航空遥感图像目标检测数据集,给出了典型算法的实验比较结果,最后给出现阶段航空遥感图像目标检测研究中所存在的问题,并对未来研究及发展趋势进行了展望。     

Oil palm plantation mapping from high-resolution remote sensing images using deep learning

ABSTRACT

Oil palm plantation mapping is an important task in land planning and management in Malaysia. Most existing studies were based on satellite images using traditional machine learning or image segmentation methods. In order to obtain finer oil palm plantation maps from high spatial-resolution satellite images, we proposed a novel deep learning-based semantic segmentation approach, named Residual Channel Attention Network (RCANet). It consists of an encoder-decoder architecture and a post-processing component. The Residual Channel Attention Unit (RCAU) designed in our proposed approach reuses the low-level features extracted from the encoder part through upsampling, effectively enhancing the discriminative features and suppressing the indiscriminate features. We extended the fully connected Conditional Random Field (FC-CRF) in the post-processing to further refine the segmentation results. Experiment results were evaluated by our proposed Malaysian Oil Palm Plantation Dataset (MOPPD), which was collected from the Google Earth high spatial-resolution image and published in this article. Our proposed method achieves the overall accuracy (OA) of 96.88% and mean Intersection-over-Union (mean IoU) of 90.58%, improving the OA by 2.03%-3.96% and the mean IoU by 2.13%-5.44% compared with other semantic segmentation methods (i.e. Fully Connected Network, U-Net and Fully Connected DenseNet). In addition, we exhibited the results of the oil palm plantation mapping in large-scale areas (around 320 km²) and demonstrated the effectiveness of our method for large-scale mapping.     

A new change-detection method in high-resolution remote sensing images based on a conditional random field model

A new change-detection method for remote sensing images based on a conditional random field (CRF) model is proposed in this paper. The method artfully uses memberships of Fuzzy C-means as unary potentials in the fully connected CRF (FCCRF) model without training parameters, and pairwise potentials of the CRF model are defined by a linear combination of Gaussian kernels, with which a highly efficient approximate inference algorithm can be used. The proposed FCCRF model is expressed on the complete set of pixels in both the observed multitemporal images, which can incorporate long range contextual information of remote-sensing images and enable greatly refined change-detection results. Experimental results demonstrate that the proposed approach leads to more accurate pixel-level change-detection performance and is more robust against noise than traditional algorithms.     

跨层细节感知和分组注意力引导的遥感图像语义分割

目的 语义分割是遥感图像智能解译的关键任务之一,遥感图像覆盖面广,背景交叉复杂,且地物尺寸差异性大。现有方法在复杂背景下的多尺度地物上分割效果较差,且分割区域破碎边界不连续。针对上述问题,提出了一种跨层细节感知和分组注意力引导的语义分割模型用于高分辨率遥感图像解析。方法 首先采用结构新颖的ConvNeXt骨干网络,编码输入图像的各层次特征。其次,设计了分组协同注意力模块,分组并行建模通道和空间维度的特征依赖性,通道注意力和空间注意力协同强化重要通道和区域的特征信息。接着,引入了自注意力机制,构建了跨层细节感知模块,利用低层特征中丰富的细节信息,指导高层特征层学习空间细节,保证分割结果的区域完整性和边界连续性。最后,以山西省太原市为研究区域,自制高分辨率遥感太原市城区土地覆盖数据集（Taiyuan urban land cover dataset,TULCD）,所提方法实现了太原市城区土地覆盖精细分类任务。结果 实验在自制数据集TULCD和公开数据集Vaihingen上与最新的5种算法进行了比较,所提方法在两个数据集上平均像素准确率（mean pixel accuracy,mPA）为74.23%、87.26%,平均交并比（mean intersection over union,mIoU）为58.91%、77.02%,平均得分mF1为72.24%、86.35%,均优于对比算法。结论 本文提出的高分辨率遥感图像语义分割模型具有较强的空间和细节感知能力,对类间差异小的相邻地物也有较强的鉴别能力,模型的整体分割精度较高。     

A novel method for vehicle detection in high-resolution aerial remote sensing images using YOLT approach

Multimedia Tools and Applications - Owing to the innovative challenge stood by an intergovernmental military alliance, we have proposed a model to find novel solutions in the areas of data...     

Modeling and simulation for natural disaster contingency planning driven by high-resolution remote sensing images

Natural disasters occur unexpectedly and usually result in huge losses of life and property. How to effectively make contingency plans is an intriguing question constantly faced by governments and experts. Human rescue operations are the most critical issue in contingency planning. A natural disaster scenario is, in general, highly complicated and dynamic. Modeling and simulation technologies have been gaining considerable momentum in investigating natural disaster scenarios to enable contingency planning. However, existing M&S systems still suffer from two open problems: (1) a lack of real data on natural disasters; and (2) the absence of methods and platforms to describe the collective behaviors of people in disaster situations. Considering these problems, an M&S framework for human rescue operations in a typical natural disaster, i.e., a landslide, has been developed in this study. The framework consists of three modules: (1) remote sensing information extraction, (2) landslide simulation, and (3) crowd simulation. The crowd simulation module is driven by the real/virtual data provided by the former modules. A number of simulations (using the Zhouqu landslide as an example) have been performed to study human relief operations spontaneously and under manipulation, with the effect of contingency plans highlighted. The experimental results demonstrate that  (1) the simulation framework is an effective tool for contingency planning, and (2) real data can make the simulation outputs more meaningful.     

Study of land cover classification based on knowledge rules using high-resolution remote sensing images

This paper deals with the limitations of visual interpretation of high-resolution remote sensing images and of automatic computer classification completely dependent on spectral data. A knowledge-rule method is proposed, based on spectral features, texture features obtained from the gray-level co-occurrence matrix, and shape features. QuickBird remote sensing data were used for an experimental study of land-use classification in the combination zone between urban and suburban areas in Beijing. The results show that the deficiencies of methods where only spectral data are used for classification can be eliminated, the problem of similar spectra in multispectral images can be effectively solved for the classification of ground objects, and relatively high classification accuracy can be reached.     

改进卷积网络的高分遥感图像城镇建成区提取

目的 城镇建成区是城镇研究重要的基础信息,也是实施区域规划、落实城镇功能空间布局的前提。但是遥感影像中城镇建成区的环境复杂,同时不同城镇建成区在坐落位置、发展规模等方面存在许多差异,导致其信息提取存在一定困难。方法 本文基于面向图像语义分割的深度卷积神经网络,使用针对特征图的强化模块和通道域的注意力模块,对原始DeepLab网络进行改进,并通过滑动窗口预测、全连接条件随机场处理方法,更准确地实现城镇建成区提取。同时,针对使用深度学习算法容易出现过拟合和鲁棒性不强的问题,采用数据扩充增强技术进一步提升模型能力。结果 实验数据是三亚和海口部分地区的高分二号遥感影像。结果表明,本文方法的正确率高于93%,Kappa系数大于0.837,可以有效地提取出大尺度高分辨率遥感影像中的城镇建成区,且提取结果最为接近实际情况。结论 针对高分辨率遥感卫星影像中城镇建成区的光谱信息多样化、纹理结构复杂化等特点,本文算法能在特征提取网络中获取更多特征信息。本文使用改进的深度学习方法,提出两种处理方法,显著提高了模型的精度,在实际大幅遥感影像的使用中表现优秀,具有重要的实用价值和广阔的应用前景。     

结合非监督分类和几何—纹理—光谱特征的高分影像道路提取

目的 目前针对复杂场景高分辨率遥感影像道路提取多采用监督分类方法,但需要人工选择样本,自动化程度低且具有不稳定性。基于像元级的方法,提取完整度低且易产生椒盐噪声;面向对象的方法易产生粘连问题。为了提高道路提取的完整度、准确度和自动化程度,提出一种基于非监督分类和几何—纹理—光谱特征的道路提取方法。方法 首先考虑光谱特征利用非监督分类进行初步分割,结合基于纹理特征分类的结果得到初始道路区域。然后根据道路特征建立一套完整的非道路区域滤除体系：边缘滤波断开道路和非道路的连接、纹理滤波滤除大面积非道路区域、形状滤波去除剩余小面积非道路区域。最后利用张量投票算法得到连贯、平滑的道路中心线。结果 选择复杂场景下的高分辨率IKONOS影像和QuickBird影像进行实验,与国内外基于像素和面向对象的两种有代表性的道路提取方法进行对比,采用完整率、正确率、检测质量3个评价指标进行定量评价。实验结果表明该方法相比于其他算法在完整率、正确率和检测质量上平均提高26.61%、5.57%和26.77%。定性分析结果表明,本文方法可以有效改善椒盐噪声和粘连现象。此外本文方法的自动化程度更高。结论 提出了一种基于非监督分类和几何—纹理—光谱特征的高分辨遥感影像道路提取方法,非监督相对于监督分类的方法有更高的自动化程度,复杂场景下的道路提取融合几何—纹理—光谱特征有效避免了基于像元级道路提取易产生的椒盐噪声现象和面向对象道路提取易产生的粘连现象。该方法适用于高分辨率遥感影像城市道路提取,能够得到较高的完整度、准确度以及自动化程度。非监督分类和多特征结合的道路提取方法有广阔的应用前景。     

GSSAC:一种用于遥感影像配准的误匹配点检测方法*

遥感影像配准中,由于光照、成像角度、几何变形等因素的影响,无论采用何种配准方法,总会产生误匹配点,因此误匹配点检测也是一个非常重要的步骤。针对常用RANSAC (Random Sample Consensus)方法不稳定、无法准确检测分布不均匀匹配点的缺点,提出了分组排序采样一致性(Group Sorted Sample Consensus, GSSAC)方法来提高误匹配点检测的稳定性和精度。分组排序采样方法首先将匹配点分为若干组,在每组内计算匹配点的误差并排序,然后在每组中分别采样若干个匹配点组成估算模型参数需要的匹配点。实验结果表明,GSSAC方法可以稳定的获得高精度的检测结果。     

遥感图像变化区域的无监督压缩感知

传统的基于结构特征的遥感图像变化检测方法,易受成像稳定性的影响而误差很大。针对图像内在的稀疏性结构信息,提出基于压缩感知(CS)的遥感图像变化检测方法。通过自适应构造超完备字典将图像局部信息投影到高维空间中,实现图像的稀疏表示,并运用随机矩阵得到了数据在高维空间中的低维特征子空间。最后利用模糊C均值(FCM)聚类算法进行无监督聚类,实现遥感图像变化区域信息的重构。实验结果表明,本文方法不仅能够很好的检测出图像的轮廓变化和图像的区域变化,而且对噪声具有很好的鲁棒性。     

Dempster-Shafer证据融合形状特征的高分辨率遥感图像道路信息提取

道路提取是遥感图像目标识别和提取中一项具有重要意义而困难的任务。在遥感图像道路提取的过程中,由于道路的不同形状和图像信息的复杂性,目前在许多基于形状特征提取道路的方法中,选取形状特征阈值时具有一定的难度,且需要大量的人工干预操作,缺乏一定的通用性,因此,本文提出一种基于DS(dempster-shafer)证据理论和形状特征的道路提取方法。该方法首先对道路的几何形状特征进行分析和优化,据此设计概率分配函数,并利用DS证据理论融合形状特征以获取道路段,最后通过道路连接操作得到道路的中心线。文末通过对典型道路图像和非典型道路图像的实验表明,该方法能够降低选取形状特征阈值的难度和对人工的依赖性,能适用于高分辨率遥感图像中直线型和曲线型道路的提取,具有一定的可行性。