A probabilistic framework for improving reverse geocoding output

Reverse geocoding, which transforms machine‐readable GPS coordinates into human‐readable location information, is widely used in a variety of location‐based services and analysis. The output quality of reverse geocoding is critical because it can greatly impact these services provided to end‐users. We argue that the output of reverse geocoding should be spatially close to and topologically correct with respect to the input coordinates, contain multiple suggestions ranked by a uniform standard, and incorporate GPS uncertainties. However, existing reverse geocoding systems often fail to fulfill these aims. To further improve the reverse geocoding process, we propose a probabilistic framework that includes: (1) a new workflow that can adapt all existing address models and unitizes distance and topology relations among retrieved reference data for candidate selections; (2) an advanced scoring mechanism that quantifies characteristics of the entire workflow and orders candidates according to their likelihood of being the best candidate; and (3) a novel algorithm that derives statistical surfaces for input GPS uncertainties and propagates such uncertainties into final output lists. The efficiency of the proposed approaches is demonstrated through comparisons to the four commercial reverse geocoding systems and through human judgments. We envision that more advanced reverse geocoding output ranking algorithms specific to different application scenarios can be built upon this work.     

基于深度学习的GPS水准拟合方法

针对数学拟合法在进行全球定位系统(GPS)水准拟合时,因受自身模型限制,导致GPS水准拟合精度不高的问题,该文提出了一种基于EGM2008模型和深度学习的GPS水准拟合法.首先使用深度学习中的分段线性整流函数(ReLU)作为神经元激活函数加快网络的收敛速度,然后利用 自适应矩估计函数(Adam)作为优化函数加速获取最优解,并采用正则化丢弃法(Dropout)增强深度学习网络的泛化能力.通过实测数据计算表明:该文方法相比常用的多项式拟合法,丘陵地区外符合精度提高了约65％,达到1.7 cm;高差变化较大的山地外符合精度提高了约90％,达到1.2 cm.     

基于深度学习的人群活动流量时空预测模型

传统的时空预测方法缺乏对复杂时空非线性关系的描述,且难以顾及空间多尺度特征对于预测结果的影响.针对这一问题,本文提出了一种融合空间多尺度特征的时空网络模型(MST-Net),将流量预测的回归问题转换为具有时空特性的判别模型.首先,通过并联卷积提取空间多尺度特征;然后,通过引入注意力机制的门控循环单元提取时间特征;最后,...     

Multi-unit building address geocoding: An approach without indoor location reference data

Accurately mapped locations within multi-unit properties are useful for several organizations in today's society. Published work on geocoding methods either require detailed location reference data or does not apply to multi-unit buildings. In this research, a generalizable method is realized to map apartment addresses to their explicit locations without access to indoor location reference data based on publicly available address- and geospatial-building information. The performance of this approach is measured by conducting a comparative study between a linear interpolation baseline and gradient-boosted decision trees model. The proposed method can successfully geocode addresses across different building shapes and sizes. Furthermore, the model significantly outperforms the baseline in terms of positional accuracy proving the feasibility of approximating apartment locations by their address- and geospatial-building information.     

基于深度学习的滑坡位移时空预测

滑坡变形监测数据是认识滑坡变形演化规律的直接依据,对该数据深度挖掘是实现滑坡灾害预警预报的有力保障。现有的滑坡位移预测模型多局限于单个监测点的时序预测,且未考虑监测点间的空间相关性。针对上述问题,本文提出了一种基于深度学习的滑坡位移时空预测模型:首先,构建表达所有点间空间相关性的加权邻接矩阵;其次,引入外界影响因素加强属性特征矩阵,以构建图结构数据;最后,采用集合图卷积网络(GCN)和门控循环单元(GRU)的深度学习模型,并通过多组试验寻找最优超参数,实现滑坡位移的时空预测。本文模型结果的均方根误差为4.429 mm,与对比模型相比至少降低了47.3%。而消融试验结果也显示,引入外界影响因素的属性增强可进一步提高模型的预测性能,均方根误差相对于未属性增强结果减少了28.4%。结果表明,该方法可用于滑坡位移或其他地质灾害中同样具有时空关联属性的观测量的时空预测。     

融合城市行道树特征选取模型的自适应深度学习分类

针对城市行道树的学习多分类问题,本文在综合分析城市行道树多分类特征的基础上,提出一种融合特征自动选取模型的自适应深度学习方法。基于随机森林法,学习行道树的特征重要性,通过特征消除方法舍弃不重要的特征,实现城市行道树多分类特征自动选取;在城市行道树分类特征工程提取的基础上,构建了城市行道树多分类问题的自适应深度学习方法,并采用交叉验证与参数搜索方法,对所提出的深度学习模型进行改进。试验结果表明,本文所提出的融合特征自动选取模型的自适应深度学习方法具有良好性能,解决了城市行道树多分类预测的准确性与泛化问题。     

适用于城市场景大规模点云语义标识的深度学习网络

近年来,以点云为代表的三维数据不断涌现,如何利用人工智能手段,高度提升点云的解译能力,实现城市地物目标的语义标识、三维精准提取等成为亟待攻克的难题.为此,本文提出了一种端到端的三维点云深度学习网络,通过构建不规则分布点云的上下采样策略、特征多层聚合与传播,以及顾及样本不均的损失函数,有效保障了点云采样的高效性、特征提取...     

点云深度学习基准数据集

为推进深度学习方法在点云配准、语义分割、实例分割等领域的发展,武汉大学联合国内外多家高等院校和研究机构发布了包含多类型场景的地面站点云配准基准数据集WHU-TLS和包含语义、实例的城市级车载点云基准数据集WHU-MLS.其中,WHU-TLS基准数据集涵盖了地铁站、高铁站、山地、公园、校园、住宅、河岸、文化遗产建筑、地下...     

高光谱遥感影像多级联森林深度网络分类算法

高光谱遥感技术在环境监测、应急保障、精细地物提取等方面有着广泛的应用,随着高分五号高光谱数据的正式发布,高光谱遥感技术将发挥更重要的作用。遥感影像分类作为高光谱遥感影像信息处理的重要部分,已成为当前研究重点。本文针对传统多级联森林深度学习中模型复杂、无法利用基分类器差异信息、对类间差异较小的样本无法正确区分等不足,提出了一种改进的多级联森林深度学习模型,在模型框架中,分别采用了随机森林和旋转森林作为基分类器,并引入逻辑回归分类器作为判别器用于训练层扩展。相较于传统的深度神经网络,改进的多级联森林深度网络超参数较少且能够自适应确定训练层,更方便进行模型优化。实验采用了高分五号数据集及两个公开的高光谱数据集(Indian Pines数据集及Pavia University数据集)进行精度评定,同时选择了传统分类器支持向量机、深度置信网等模型作为对比分析。实验结果表明,改进的多级联森林深度学习模型能有效地进行高光谱遥感影像分类,且较传统的分类方法精度有所提升。     

基于深度学习的高分辨率遥感影像建筑物提取方法

针对传统的建筑物提取方法精度较低和边界不完整等问题,本文提出基于深度学习的高分辨率遥感影像建筑物提取方法。首先,采用主成分变换非监督预训练网络结构,获得待提取遥感影像特征。其次,为减少在池化过程中影像特征信息的丢失,提出自适应池化模型,通过非下采样轮廓波变换来获取影像纹理特征,并将纹理特征输入网络中参与建筑物提取。最后,将影像特征输入softmax分类器进行分类,获得建筑物提取结果。选取典型区域进行建筑物提取试验,并与典型建筑物提取方法进行对比分析,结果表明,本文提取方法精度高,并且提取建筑物的边界清晰、完整。     

Local motion simulation using deep reinforcement learning

Traditional local motion simulation focuses largely on avoiding collisions in the next frame. However, due to its lack of forward looking, the global trajectory of agents usually seems unreasonable. As a method of optimizing the overall reward, deep reinforcement learning (DRL) can better correct the problems that exist in the traditional local motion simulation method. In this article, we propose a local motion simulation method integrating optimal reciprocal collision avoidance (ORCA) and DRL, referred to as ORCA‐DRL. The main idea of ORCA‐DRL is to perform local collision avoidance detection via ORCA and smooth the trajectory at the same time via DRL. We use a deep neural network (DNN) as the state‐to‐action mapping function, where the state information is detected by virtual visual sensors and the action space includes two continuous spaces: speed and direction. To improve data utilization and speed up the training process, we use the proximal policy optimization based on the actor–critic (AC) framework to update the DNN parameters. Three scenes (circle, hallway, and crossing) are designed to evaluate the performance of ORCA‐DRL. The results reveal that, compared with the ORCA, our proposed ORCA‐DRL method can: (a) reduce the total number of frames, leading to less time for agents to reach their destination; and (b) effectively avoid local optima, evidenced by smoothed global trajectories.     

顾及局部特征的深度学习点云分类研究

针对PointNet深度学习算法可以直接处理无序点云并取得了良好的精度,但是缺乏对局部信息学习过程的问题,该文基于图卷积模型,在PointNet基础上构造层次化的K邻域图,扩大局部感受野,获得高层次的特征抽象,有效提取了点云的局部特征从而提高了分类精度.分类实验在ModelNet40数据集上进行,取得了91.2％的测试精度.研究结果表明,该文提出的算法比PointNet分类结果高出2.0％,同时本文构造的分类网络鲁棒性优于PointNet算法,为点云分类工作提供了一种有效思路.     

深度学习赋能地图制图的若干思考

地图制图学包含地图制作与地图应用两大任务,分别与人工智能技术有不解之缘.经历了符号主义智能表达的地图制图专家系统、行为主义智能表达的空间优化决策后,地图制图面临与连接主义下的深度学习的结合,以提升地图制图的智能化水平.本文针对"深度学习+地图制图"命题讨论了3个问题.一是从深度学习方法与地图空间问题解答策略思想的一致性,基于梯度下降、局部相关、特征降维和非线性化性质,回答了两者结合的可行性;二是从地图学独特的学科特点和技术环境分析了两者结合面临的挑战,涉及地图数据组织的非规范性、样本建立的专业需求、几何与地理特征的融合,以及地图固有的空间尺度性;三是分别讨论了地图制作与地图应用融入深度学习的切入点和具体方法.     

A multi-index learning approach for classification of high-resolution remotely sensed images over urban areas

In recent years, it has been widely agreed that spatial features derived from textural, structural, and object-based methods are important information sources to complement spectral properties for accurate urban classification of high-resolution imagery. However, the spatial features always refer to a series of parameters, such as scales, directions, and statistical measures, leading to high-dimensional feature space. The high-dimensional space is almost impractical to deal with considering the huge storage and computational cost while processing high-resolution images. To this aim, we propose a novel multi-index learning (MIL) method, where a set of low-dimensional information indices is used to represent the complex geospatial scenes in high-resolution images. Specifically, two categories of indices are proposed in the study: (1) Primitive indices (PI): High-resolution urban scenes are represented using a group of primitives (e.g., building/shadow/vegetation) that are calculated automatically and rapidly; (2) Variation indices (VI): A couple of spectral and spatial variation indices are proposed based on the 3D wavelet transformation in order to describe the local variation in the joint spectral-spatial domains. In this way, urban landscapes can be decomposed into a set of low-dimensional and semantic indices replacing the high-dimensional but low-level features (e.g., textures). The information indices are then learned via the multi-kernel support vector machines. The proposed MIL method is evaluated using various high-resolution images including GeoEye-1, QuickBird, WorldView-2, and ZY-3, as well as an elaborate comparison to the state-of-the-art image classification algorithms such as object-based analysis, and spectral-spatial approaches based on textural and morphological features. It is revealed that the MIL method is able to achieve promising results with a low-dimensional feature space, and, provide a practical strategy for processing large-scale high-resolution images.     

AMSR-E地表温度数据重建深度学习方法

地表温度对于全球气候变化等研究具有重要意义。被动微波遥感传感器AMSR-E (Advanced Microwave Scanning Radiometer for EOS)可以获得全天候地表温度,可作为多云条件下热红外地表温度数据的补充;但轨道扫描间隙限制了该数据在全球或区域尺度上的实际应用。鉴于地表温度的高时空异质性和AMSR-E LST轨道间隙数据的特点,本文提出了一种多时相特征连接卷积神经网络地表温度双向重建模型（MTFC-CNN）,利用深度学习在处理复杂非线性问题上的优势,重建轨道间隙区域的地表温度值。将2010年中国大陆四季的AMSR-E LST数据（数据未含港澳台区域）,分为白天和夜晚,形成共8个数据子集进行实验。在模拟实验中,重建结果与原始反演地表温度值平均均方根误差在1.0 K左右,决定系数R2在0.88以上,优于传统的样条空间插值和时间线性回归方法;真实实验结果具有较好的目视效果,且与对应MODIS LST产品对比发现,重建区LST值和未重建区LST值与MODIS LST产品间具有相近的平均均方根误差和决定系数。因此,本文提出的MTFC-CNN方法能有效重建AMSR-...     

智能遥感深度学习框架与模型设计EI北大核心CSCD

近年来,随着遥感技术的快速发展,遥感对地观测数据获取量与日俱增。在对海量遥感数据的特征提取与表征上,基于深度学习的智能遥感影像解译技术展现出了显著优势。然而,遥感影像智能处理框架和信息服务能力还相对滞后,开源的深度学习框架与模型尚不能满足遥感智能处理的需求。在分析现有深度学习框架和模型的基础上,针对遥感影像幅面大、尺度变化大、数据通道多等问题,本文设计了嵌入遥感特性的专用深度学习框架,并重点讨论了其构建方法,以及地物分类任务的初步试验结果等。本文提出的智能遥感解译框架架构将为构建具备多维时空谱遥感特性的深度学习框架与模型提供有力支撑。     

Satellite selection with an end-to-end deep learning network

Benefiting from multi-constellation Global Navigation Satellite Systems (GNSS), more and more visible satellites can be used to improve user positioning performance. However, due to limited tracking receiver channels and power consumption, and other issues, it may be not possible, or desirable, to use all satellites in view for positioning. The optimal subset is generally selected from all possible satellite combinations to minimize either Geometric Dilution of Precision (GDOP) or weighted GDOP (WGDOP). However, this brute force approach is difficult to implement in real-time applications due to the time- and power-consuming calculation of the DOP values. As an alternative to a brute force satellite selection procedure, the authors propose an end-to-end deep learning network for satellite selection based on the PointNet and VoxelNet networks. The satellite selection is converted to a satellite segmentation problem, with specified input channel for each satellite and two class labels, one for selected satellites and the other for those not selected. The aim of the satellite segmentation is that a fixed number of satellites with the minimum GDOP/WGDOP value can be segmented from any feeding order of input satellites. To validate the proposed satellite segmentation network, training and test data from 220 IGS stations tracking GPS and GLONASS satellites were used. The segmentation performance using different architectures and representations of input channels, including receiver-to-satellite unit vector and elevation and azimuth, were compared. It was found that the input channel with elevation and azimuth can achieve better performance than using the receiver-to-satellite unit vector, and an architecture with stacked feature encoding (FE) layers has better satellite segmentation performance than one without stacked FE layers. In addition, the models with GDOP and WGDOP criteria for selecting 9 and 12 satellites were trained. It was demonstrated that the satellite segmentation network was about 90 times faster than using the brute force approach. Furthermore, all the trained models can effectively select the satellites making the most contribution to the desired GDOP/WGDOP value. Approximately 99% of the tests had GDOP and WGDOP value differences smaller than 0.03 and 0.2, respectively, between the predicted subset and the optimal subset.     

深度学习农作物分类的弱样本适用性

基于大数据驱动的深度学习挖掘图像数据的规律和层次已成为遥感影像解译的研究热点。海量标签样本是训练深度学习模型的前提条件,但成本昂贵的人工标记样本限制了深度学习技术在遥感领域的应用。本文提出了一种基于弱样本的深度学习模型农作物分类策略：以GF-1影像为数据源,将传统分类器SVM分类结果视为弱样本,训练深度卷积网络模型DCNN (Deep Convolutional Neural Networks),获取辽宁省水稻和玉米的空间分布,分析弱样本的适用性。结果显示：测试集总体精度达到0.90,水稻和玉米F1分数分别为0.81和0.90;在不同地形地貌、复杂种植结构的农业景观下均表现出良好的分类效果;与SVM结果的空间一致性为0.90;当弱样本最大面积误差比例小于0.36时,弱样本仍适用于DCNN作物分类,结果的总体精度保持在0.86以上。综上,该策略一定程度上消除了深度学习模型对大量人工标记样本高度依赖的局限性,为实现大尺度农作物遥感分类提供了一种新途径。     

地图空间形状认知的自编码器深度学习方法

形状是地理空间要素的重要特征,是人们建立空间概念、形成空间认知的重要依据.本文利用深度学习的特征挖掘能力引入自编码学习方法,对二维地图空间中形状边界上多组邻域尺寸下的多个特征进行集成和整合,为空间形状认知的机理和形式化提供支撑.本文以建筑物数据为例,将建筑物形状边界转换为序列数据,并提取其描述特征;随后结合sequence-to-sequence自编码学习模型,对无标签的建筑面要素数据进行学习训练,形成形状认知编码.试验表明,本文方法能够产生符合形状认知、具有相似度计算意义的形状编码,具备对不同建筑物形状的区分能力;同时,在形状检索和匹配等应用场景中,该形状编码能有效地表示建筑物的全局和局部特征,与视觉认知结果一致.     

基于自主学习的地图学PBL教学方法研究

研究在地图学课程教学过程中,将传统教学方法和PBL教学方法相结合,设计以任务为驱动的自主学习环节,并通过实证分析学牛对不同教学方法的接受程度、在PBL教学方法中所存在问题以及适宜于PBL教学模式的考评方式.研究表明,地图学教学应改变原有的教学模式,在传统教学模式的基础上,结合PBL教学方法逐步开展.