基于ImageNet预训练卷积神经网络的遥感图像检索

高分辨率遥感图像内容复杂,细节信息丰富,传统的浅层特征在描述这类图像上存在一定难度,容易导致检索中存在较大的语义鸿沟。本文将大规模数据集ImageNet上预训练的4种不同卷积神经网络用于遥感图像检索,首先分别提取4种网络中不同层次的输出值作为高层特征,再对高层特征进行高斯归一化,然后采用欧氏距离作为相似性度量进行检索。在UC-Merced和WHU-RS数据集上的一系列实验结果表明,4种卷积神经网络的高层特征中,以CNN-M特征的检索性能最好;与视觉词袋和全局形态纹理描述子这两种浅层特征相比,高层特征的检索平均准确率提高了15.7%~25.6%,平均归一化修改检索等级减少了17%~22.1%。因此将ImageNet上预训练的卷积神经网络用于遥感图像检索是一种有效的方法。     

2017年“8.8”九寨沟地震滑坡自动识别与空间分布特征

2017年8月8日发生的7.0级九寨沟地震诱发九寨沟熊猫海附近产生大量的滑坡体，造成道路阻塞，严重影响地震应急救援进度。为快速准确地识别滑坡分布范围，本文在深入分析滑坡遥感影像特征的基础上，引入面向对象分析方法，实现了基于无人机影像的震后滑坡体的自动识别。通过多尺度分割算法获取滑坡多层次影像对象，利用SEaTH算法自动构建每一层次特征规则集，实现基于不同层次分析的滑坡体自动识别。分析滑坡体在地形、活动断层等因子中的空间分布特征，为地震滑坡预测与危险性评价奠定基础。与人工目视解译结果相比较，基于面向对象的滑坡自动识别方法提取精度可达94.8%，Kappa系数为0.827，在电脑配置相同的情况下，自动识别方法的效率是人工目视解译效率的一倍。空间分布特征分析表明，地震滑坡的空间分布与斜坡坡度、地形起伏度呈正相关关系，与地表粗糙度存在负相关关系，研究区滑坡体分布存在明显的断层效应。     

Combining machine-learning topic models and spatiotemporal analysis of social media data for disaster footprint and damage assessment

Current disaster management procedures to cope with human and economic losses and to manage a disaster’s aftermath suffer from a number of shortcomings like high temporal lags or limited temporal and spatial resolution. This paper presents an approach to analyze social media posts to assess the footprint of and the damage caused by natural disasters through combining machine-learning techniques (Latent Dirichlet Allocation) for semantic information extraction with spatial and temporal analysis (local spatial autocorrelation) for hot spot detection. Our results demonstrate that earthquake footprints can be reliably and accurately identified in our use case. More, a number of relevant semantic topics can be automatically identified without a priori knowledge, revealing clearly differing temporal and spatial signatures. Furthermore, we are able to generate a damage map that indicates where significant losses have occurred. The validation of our results using statistical measures, complemented by the official earthquake footprint by US Geological Survey and the results of the HAZUS loss model, shows that our approach produces valid and reliable outputs. Thus, our approach may improve current disaster management procedures through generating a new and unseen information layer in near real time.     

A multi-level context-guided classification method with object-based convolutional neural network for land cover classification using very high resolution remote sensing images

Classification of very high resolution imagery (VHRI) is challenging due to the difficulty in mining complex spatial and spectral patterns from rich image details. Various object-based Convolutional Neural Networks (OCNN) for VHRI classification have been proposed to overcome the drawbacks of the redundant pixel-wise CNNs, owing to their low computational cost and fine contour-preserving. However, classification performance of OCNN is still limited by geometric distortions, insufficient feature representation, and lack of contextual guidance. In this paper, an innovative multi-level context-guided classification method with the OCNN (MLCG-OCNN) is proposed. A feature-fusing OCNN, including the object contour-preserving mask strategy with the supplement of object deformation coefficient, is developed for accurate object discrimination by learning simultaneously high-level features from independent spectral patterns, geometric characteristics, and object-level contextual information. Then pixel-level contextual guidance is used to further improve the per-object classification results. The MLCG-OCNN method is intentionally tested on two validated small image datasets with limited training samples, to assess the performance in applications of land cover classification where a trade-off between time-consumption of sample training and overall accuracy needs to be found, as it is very common in the practice. Compared with traditional benchmark methods including the patch-based per-pixel CNN (PBPP), the patch-based per-object CNN (PBPO), the pixel-wise CNN with object segmentation refinement (PO), semantic segmentation U-Net (U-NET), and DeepLabV3+(DLV3+), MLCG-OCNN method achieves remarkable classification performance (> 80 %). Compared with the state-of-the-art architecture DeepLabV3+, the MLCG-OCNN method demonstrates high computational efficiency for VHRI classification (4–5 times faster).     

遥感影像目标的尺度特征卷积神经网络识别法

高分辨率遥感影像的目标检测与识别，是高分对地观测系统中影像信息自动提取及分析理解的重要内容。针对传统影像目标检测与识别算法中人工设计特征稳健性与普适性差的问题，本文提出基于高分辨率遥感影像目标尺度特征的卷积神经网络检测与识别方法。首先通过统计遥感影像目标的尺度范围，获得卷积神经网络训练与测试过程中目标感兴趣区域合适的尺度大小。然后根据目标感兴趣区域合适的尺度，提出基于高分辨率遥感影像目标尺度特征的卷积神经网络检测与识别架构。通过WHU-RSone数据集对本文卷积神经网络架构与Faster-RCNN架构对比测试验证。试验结果表明，本文架构ZF模型和本文架构VGG-16模型的mean average precision （mAP）分别比Faster-RCNN ZF模型和Faster-RCNN VGG-16模型提高8.17%和8.31%，本文卷积神经网络架构可获得良好的影像目标检测与识别效果。     

基于CNN-GCN双流网络的高分辨率遥感影像场景分类

高分辨率遥感影像具有复杂的几何结构和空间布局,传统的卷积神经网络的方法仅能提取场景图像中的全局特征,忽略了上下文的关系,导致特征的表达能力受限,制约了分类精度提高。针对此问题,本文提出一个面向高分辨率遥感影像场景分类的CNN-GCN双流网络,该算法包含CNN流和GCN流两个模块。CNN流基于预训练DenseNet-121网络提取高分影像的全局特征;而GCN流采用由预训练VGGNet-16网络得到的卷积特征图构建邻接图,再通过GCN模型提取高分影像的上下文特征。最后,通过加权级联的方式有效地融合全局特征和上下文特征并利用线性分类器实现分类。本文选取AID、RSSCN7和NWPU-RESISC45共3个具有挑战性的数据集进行实验,得到的最高分类精度分别是97.14%、95.46%和94.12%,结果表明本文算法能够有效地表征场景并取得具有竞争力的分类结果。     

迁移学习支持下的高分影像山地滑坡灾害解译模型

无人机低空遥感是近年来新兴的一种快速获取灾情信息的手段,如何利用无人机高分影像构建滑坡灾害解译模型是实现快速自动解译滑坡的关键。针对该问题,对比了多种影像特征提取方法,将迁移学习(TL)特征和支持向量机(SVM)引入到构建滑坡灾害自动解译模型中,提出了一种TL支持下的高分影像滑坡灾害解译模型。选取5·12汶川地震及4·20芦山地震系列无人机影像构建了滑坡灾害样本库并进行了实验,TL特征方法整体分类准确度ACC为95%,ROC达到0.98,识别准确率达到97%。结果表明,所提方法可用于高分影像滑坡自动解译,同时可用于大面积高分影像中快速山地滑坡灾害定位及检测。     

Design and evaluation of a geovisual analytics system for uncovering patterns in spatio-temporal event data

ABSTRACT

It remains difficult to develop a clear understanding of geo-located events and their relationships to one another, particularly when it comes to identifying patterns of events in less-structured textual sources, such as news feeds and social media streams. Here we present a geovisualization tool that can leverage computational methods, such as T-pattern analysis, for extracting patterns of interest from event data streams. Our system, STempo, includes coordinated-view geovisualization components designed to support visual exploration and analysis of event data, and patterns extracted from those data, in terms of time, geography, and content. Through a user evaluation, we explore the usability and utility of STempo for understanding patterns of recent political, social, economic, and military events in Syria.     

多层特征与上下文信息相结合的光学遥感影像目标检测

目标检测是遥感影像分析的基础和关键。针对光学遥感影像中目标尺度多样、小目标居多、相似性及背景复杂等问题，本文提出一种将卷积神经网络（CNN）和混合波尔兹曼机（HRBM）相结合的遥感影像目标检测方法。首先设计细节—语义特征融合网络（D-SFN）提取卷积神经网络低层和高层融合特征，提升目标特征的判别力，特别是小目标；其次考虑上下文信息对目标检测的影响，结合上下文信息进一步加强目标表征的准确性，提升检测精度。在NWPU数据集上试验表明，本文方法能够显著提升目标检测精度且具有一定程度的稳健性。     

高分三号卫星全极化SAR影像九寨沟地震滑坡普查

基于光学遥感影像的区域滑坡普查易受云雾天气的影响,存在滑坡体调查不全面的问题,无法满足震后应急调查与恢复重建的需求。本文提出了一种极化SAR卫星数据滑坡普查方法,采用高分三号全极化SAR卫星影像数据,以九寨沟地震震区为实验区,在深入分析滑坡体和其他地物类型散射特征的基础上,融合极化特征、纹理特征和地形特征等多维特征信息,结合高分二号影像获取的训练样本,构建基于BP神经网络的全极化SAR数据滑坡自动识别模型,实现滑坡体的自动快速识别。与高分辨率光学影像与无人机航空影像目视解译结果相比较,总体识别精度为92.8%,Kappa系数为0.715,识别准确度满足地震应急实际应用的需求。研究成果可用于震区大区域滑坡体的普查,为后续开展无人机高分辨率影像滑坡体详查、灾后应急与景区恢复提供辅助信息支撑,并促进国产高分SAR卫星数据在防震减灾中的应用。     

Comparative performance of convolutional neural network,weighted and conventional support vector machine and random forest for classifying tree species using hyperspectral and photogrammetric data

ABSTRACT

The classification of tree species can significantly benefit from high spatial and spectral information acquired by unmanned aerial vehicles (UAVs) associated with advanced classification methods. This study investigated the following topics concerning the classification of 16 tree species in two subtropical forest fragments of Southern Brazil: i) the potential integration of UAV-borne hyperspectral images with 3D information derived from their photogrammetric point cloud (PPC); ii) the performance of two machine learning methods (support vector machine – SVM and random forest – RF) when employing different datasets at a pixel and individual tree crown (ITC) levels; iii) the potential of two methods for dealing with the imbalanced sample set problem: a new weighted SVM (wSVM) approach, which attributes different weights to each sample and class, and a deep learning classifier (convolutional neural network – CNN), associated with a previous step to balance the sample set; and finally, iv) the potential of this last classifier for tree species classification as compared to the above mentioned machine learning methods. Results showed that the inclusion of the PPC features to the hyperspectral data provided a great accuracy increase in tree species classification results when conventional machine learning methods were applied, between 13 and 17% depending on the classifier and the study area characteristics. When using the PPC features and the canopy height model (CHM), associated with the majority vote (MV) rule, the SVM, wSVM and RF classifiers reached accuracies similar to the CNN, which outperformed these classifiers for both areas when considering the pixel-based classifications (overall accuracy of 84.4% in Area 1, and 74.95% in Area 2). The CNN was between 22% and 26% more accurate than the SVM and RF when only the hyperspectral bands were employed. The wSVM provided a slight increase in accuracy not only for some lesser represented classes, but also some major classes in Area 2. While conventional machine learning methods are faster, they demonstrated to be less stable to changes in datasets, depending on prior segmentation and hand-engineered features to reach similar accuracies to those attained by the CNN. To date, CNNs have been barely explored for the classification of tree species, and CNN-based classifications in the literature have not dealt with hyperspectral data specifically focusing on tropical environments. This paper thus presents innovative strategies for classifying tree species in subtropical forest areas at a refined legend level, integrating UAV-borne 2D hyperspectral and 3D photogrammetric data and relying on both deep and conventional machine learning approaches.     

A novel mathematical morphology based algorithm for shoreline extraction from satellite images

Abstract

Shoreline extraction is fundamental and inevitable for several studies. Ascertaining the precise spatial location of the shoreline is crucial. Recently, the need for using remote sensing data to accomplish the complex task of automatic extraction of features, such as shoreline, has considerably increased. Automated feature extraction can drastically minimize the time and cost of data acquisition and database updating. Effective and fast approaches are essential to monitor coastline retreat and update shoreline maps. Here, we present a flexible mathematical morphology-driven approach for shoreline extraction algorithm from satellite imageries. The salient features of this work are the preservation of actual size and shape of the shorelines, run-time structuring element definition, semi-automation, faster processing, and single band adaptability. The proposed approach is tested with various sensor-driven images with low to high resolutions. Accuracy of the developed methodology has been assessed with manually prepared ground truths of the study area and compared with an existing shoreline classification approach. The proposed approach is found successful in shoreline extraction from the wide variety of satellite images based on the results drawn from visual and quantitative assessments.     

膨胀卷积与金字塔表达的深度学习模型用于农村建筑物提取

由于农村建筑物结构多样、空间分布复杂等特征,自动提取面临较多困难。针对该问题,本文提出采用膨胀卷积和金字塔池化表达的神经网络模型用于遥感影像中农村建筑物自动提取。在膨胀卷积神经网络模块中,通过改变孔尺寸的大小,获取不同感受野的特征信息;在金字塔表达方面,每个模块输入不同尺度的信息,且同时下采样的倍率也不同,获取多维的金字塔尺度特征;最终将提取的浅层及深层尺度特征信息进行融合,构建一个改进的适用于农村建筑物目标自动提取的深度学习模型。试验结果表明,与FCN-8s和DeepLab模型提取的结果相比,本文方法在农村建筑物提取中表现较好的性能,提取精度明显提高,且更好保留了目标边界细节信息,减少了噪声。     

Quantitative estimates of collective geo-tagged human activities in response to typhoon Hato using location-aware big data

ABSTRACT

Location-aware big data from social media have been widely used to quantitatively characterize natural disasters and disaster-induced losses. It is not clear how human activities collectively respond to a disaster. In this study, we examined the collective human activities in response to Typhoon Hato at multi spatial scales using aggregated location request data. We proposed a Multilevel Abrupt Changes Detection (MACD) methodological framework to detect and characterize the abrupt changes in location requests in response to Typhoon Hato. Results show that, at the grid level, most anomaly grids were located within a radius of 53?km around the typhoon trajectory. At the city level, there are significant spatial difference in terms of the human activity recovery duration (230?h on average). At the subnational level, the absolute magnitude of abrupt location request changes is strongly correlated with the typhoon-induced economic losses and the population affected.     

密集卷积残差网络的遥感图像融合

针对传统的遥感图像融合方法通常会引起光谱失真的问题和大多数基于深度学习的融合方法忽略充分利用每个卷积层信息的不足,本文结合密集连接卷积网络和残差网络的特性,提出了一个新的融合网络。该网络通过建立多个密集卷积块来充分利用卷积层的分级特征,同时块与块之间通过过渡层加快信息流动,从而最大程度地对特征进行极致利用并提取到丰富的特征。该网络应用残差学习拟合深层特征与浅层特征之间的残差,加快网络的收敛速度。实验中利用GaoFen-1（GF-1）和WorldView-2/3（WV-2/3）的多光谱图像MS （Multispectral Image）和全色图像PAN（Panchromatic Image）（MS与PAN的空间分辨率之比为4）评估本文提出方法的有效性。从视觉效果和定量评估结果两个方面来看,本文方法得到的融合结果要优于所对比的传统方法和深度学习方法,并且该网络具有鲁棒性,能够泛化到不需要预训练的其他卫星图像。本文方法通过特征的重复利用实现了光谱信息的高保真并提高了空间细节分辨能力,有利于遥感图像的应用研究。     

特征提取策略对高分辨率遥感图像场景分类性能影响的评估

高分辨率遥感图像场景分类方法主要涉及两个环节:特征提取以及特征分类,分类器的设计已经相对成熟,当前工作的重点是特征提取策略的研究。为了进一步推动特征提取策略的研究,将特征提取策略对高分辨率遥感图像场景分类性能的影响进行了定性和定量评估。首先,回顾了高分辨率遥感图像场景分类的发展历程;然后,对现有高分辨率遥感图像场景分类方法的特征提取策略进行分类总结,并从理论上将各类特征提取策略对场景分类性能的影响进行定性评估;最后,在3个规模较大的数据集上对多种特征提取策略进行实验对比,将不同特征提取策略对场景分类性能的影响和各数据集的复杂度进行定量评估。     

地面建筑物破坏状态检测的多类互信息量评估—震灾前光学图像与震灾后SAR图像的融合

利用震前可具备的光学图像与震后卫星SAR图像的信息融合,可检测与评估震灾区地表面建筑物破坏状态。本文根据震前光学图像的建筑物特征参数,通过SAR成像仿真算法,模拟各建筑物SAR图像,根据震后SAR真实图像与SAR仿真图像互信息量的比较,综合利用归一化互信息量、梯度互信息量、区域互信息量及检测与评估建筑物破坏状态。采用2010年海地地震灾前的IKONOS光学图像、灾后COSMO-SkyMed与RADARSAT-2的星载SAR图像,进行震后建筑物破坏状态检测与评估。其结果与震后GeoEye光学图像进行了验证,证明了本文方法的可行性。     

空间信息技术是防洪减灾现代化的基础

我国洪涝灾害频繁,洪灾损失严重,防洪减灾将是一项长期而艰巨的任务。在防洪减灾研究中,必须积极引进当今最新科学技术成果,以高新技术改造水利这一传统产业,以水利信息化推动水利现代化。以遥感、地理信息系统和全球导航定位系统为支撑的空间信息技术可以广泛地应用于防洪减灾中,是防洪减灾现代化的基础。本文分析了空间信息技术在防洪减灾中的作用,并对空间信息技术在防洪减灾中的应用趋势作出展望。     

基于注意力密集连接金字塔网络的新增建设用地变化检测

城市新增建设用地变化迅速频繁、场景复杂等因素导致变化检测结果出现欠分割或过分割等问题,基于此本文提出了一种融合注意力机制的密集连接金字塔网络用于城市新增建设用地变化检测。在编码阶段运用卷积注意力模型提升对变化信息的关注度,突出重要特征;采用密集连接空洞卷积空间金字塔池化模块实现多尺度特征的提取与融合,提高特征的利用率与传播效率;在解码阶段通过对提取的特征图进行上采样还原图像的空间尺度特征。试验结果表明,该方法有效改善了欠分割与过分割问题,变化检测效果更好。     

地物空间关系匹配的高分辨率遥感影像检索

从低层视觉特征与地物空间关系特征对影像内容进行描述,建立检索模板与目标影像间的相似性直方图表达,提出一种适用于高分辨率遥感影像检索的新方法。首先,利用Quin+树将大幅面原始遥感影像分解为一系列同尺寸的序列子块;然后,分别提取各子块的低层视觉特征与地物关系特征,并以子块为基元构建候选子块的特征直方图;最后,对比检索模板与候选子块间的特征直方图相似性,实现高分辨率遥感影像的检索。使用多幅多源高分辨率遥感影像进行实验,结果表明本文方法对耕地、水系、建筑物等地类的检索精度大都维持在0.8以上,且各项检索性能指标均优于已有的两种遥感图像检索算法。