基于MBR组合优化算法的多尺度面实体匹配方法

针对多尺度匹配中同名实体位置偏差较大,无法直接通过面积重叠法获得候选匹配的问题,本文提出了一种基于最小外包矩形（MBR）组合优化算法的多尺度面实体匹配方法。本文方法的基本思想是通过MBR组合优化和简要的形状特征来筛选1∶1、1∶N和M∶N候选匹配,然后构建多因子人工神经网络模型来评估候选匹配。试验选取浙江省舟山市1∶2000岛礁基础数据和1∶10 000陆地基础数据中的居民地与设施面进行匹配算法的验证。结果表明,本文方法相对于基于面积重叠-神经网络的匹配方法表现出显著的优势,对存在位置偏移的匹配数据准确率和召回率分别达到了达到96.5%,达到89.0%,且能够识别所有匹配类型。     

一种用于空间数据整合的建筑物面实体对齐方法

提出了一种用于空间数据整合的建筑物面实体对齐方法,可用来改善空间数据的位置精度。首先,采用基于最小外接矩形（minimum bounding rectangle,MBR）组合优化算法的匹配方法识别整合数据之间的同名实体;然后,提出基于几何相似性的成对约束谱匹配算法检测1:1、1:N和M:N同名实体之间的共轭点对;针对1:N和M:N匹配中不可避免存在弱对应点对和错误对应点对的问题,提出基于IGG1权重的最小二乘法来有效对齐同名实体。将所提出的方法应用于对齐较高位置精度的基础测绘地图数据和较低位置精度的谷歌地图数据中,结果表明,该方法不仅可检测存在复杂轮廓对应的1:N和M:N同名实体的共轭点对,而且可实现它们之间的有效对齐,使同名实体的位置信息差异最小化。     

基于邻近模式的多比例尺居民地松弛迭代匹配

空间目标匹配是实现多源空间信息融合、空间对象变化检测与动态更新的重要前提。针对多比例尺居民地匹配问题,提出了一种基于邻近模式的松弛迭代匹配方法。该方法首先利用缓冲区分析与空间邻近关系检测候选匹配目标与邻近模式,同时计算候选匹配目标或邻近模式间的几何相似性得到初始匹配概率矩阵;然后对邻近候选匹配对进行上下文兼容性建模,利用松弛迭代方法求解多比例尺居民地的最优匹配模型,选取匹配概率最大并满足上下文一致的候选匹配目标或邻近模式为最终匹配结果。实验结果表明,所提出的多比例尺居民地匹配方法具有较高的匹配精度,能有效克服形状轮廓同质化与非均匀性偏差问题,并准确识别1:M、M:N的复杂匹配关系。     

Line segment confidence region-based string matching method for map conflation

In this paper, a method to detect corresponding point pairs between polygon object pairs with a string matching method based on a confidence region model of a line segment is proposed. The optimal point edit sequence to convert the contour of a target object into that of a reference object was found by the string matching method which minimizes its total error cost, and the corresponding point pairs were derived from the edit sequence. Because a significant amount of apparent positional discrepancies between corresponding objects are caused by spatial uncertainty and their confidence region models of line segments are therefore used in the above matching process, the proposed method obtained a high F-measure for finding matching pairs. We applied this method for built-up area polygon objects in a cadastral map and a topographical map. Regardless of their different mapping and representation rules and spatial uncertainties, the proposed method with a confidence level at 0.95 showed a matching result with an F-measure of 0.894.     

A multi‐criteria decision‐making approach for geometric matching of areal objects

We propose a method for geometric areal object matching based on multi‐criteria decision making. To enable this method, we focused on determining the matched areal object pairs that have all relations, one‐to‐one relationships to many‐to‐many relationships, in different spatial data sets by fusing geometric criteria without user invention. First, we identified candidate corresponding areal object pairs with a graph‐based approach in training data. Second, three matching criteria (areal hausdorff distance, intersection ratio, and turning function distance) were calculated in candidate corresponding pairs and these criteria were normalized. Third, the shape similarity was calculated by weighted linear combination using the normalized matching criteria (similarities) with the criteria importance through intercriteria correlation method. Fourth, a threshold (0.738) of the shape similarity estimated in the plot of precision versus recall versus all possible thresholds of training data was applied, and the matched pairs were determined and identified. Finally, we visually validated the detection of similar areal feature pairs and conducted statistical evaluation using precision, recall, and F‐measure values from a confusion matrix. Their values were 0.905, 0.848, and 0.876, respectively. These results validate that the proposed classifier, which detects 87.6% of matched areal pairs, is highly accurate.     

Improving Spatial Accuracy of Roadway Networks and Geocoded Addresses

Exposure to traffic‐related pollutants is associated with both morbidity and mortality. Because vehicle‐exhaust are highly localized, within a few hundred meters of heavily traveled roadways, highly accurate spatial data are critical in studies concerned with exposure to vehicle emissions. We compared the positional accuracy of a widely used U.S. Geological Survey (USGS) roadway network containing traffic activity data versus a global positioning system (GPS)‐validated road network without traffic information; developed a geographical information system (GIS)‐based methodology for producing improved roadway data associated with traffic activities; evaluated errors from geocoding processes; and used the CALINE4 dispersion model to demonstrate potential exposure misclassifications due to inaccurate roadway data or incorrectly geocoded addresses. The GIS‐based algorithm we developed was effective in transferring vehicle activity information from the less accurate USGS roadway network to a GPS‐accurate road network, with a match rate exceeding 95%. Large discrepancies, up to hundreds of meters, were found between the two roadway networks, with the GPS‐validated network having higher spatial accuracy. In addition, identifying and correcting errors associated with geocoding resulted in improved address matching. We demonstrated that discrepancies in roadway geometry and geocoding errors, can lead to serious exposure misclassifications, up to an order of magnitude in assigned pollutant concentrations.     

一种基于Fréchet距离的复杂线状要素匹配方法

在不同空间数据集中,同名实体往往有不同的空间表现形式,识别多源异构数据集中的同名实体是空间数据集成和应用的关键。集成不同来源的空间数据是提高GIS数据质量的重要方法,识别同名实体是数据集成和分析的先决条件。根据线要素的形状将其分为简单线要素和复杂线要素,针对现有复杂线要素匹配方法中的不足,提出了Fréchet距离的复杂线状要素匹配方法。该方法首先通过曲线要素的几何和拓扑特性获取候选匹配集,然后结合基于Fréchet距离和要素简化方法实现要素的简化。最后提出基于Fréchet距离的要素匹配改进方法,通过引入简化要素的三元组信息来存储简化后的复杂线要素的属性信息,再根据三元组信息选取要素间的匹配对,完成对不同类型匹配对的检测,实现复杂线状要素匹配。试验结果表明,该匹配方法能有效解决复杂线要素的匹配问题,并能够识别1:0、1:N和M:N匹配。     

图像匹配的物体空间关系推理表达

物体空间关系指的是物体在欧氏空间中的邻近关系,根据图像中包含物体的邻近关系解决图像匹配的问题。本文首先基于对比机制训练物体块特征提取网络,构建物体块深度特征,该特征可以有效匹配不同图像中的相同物体块;其次,基于已有的先验图像数据推理表达图像中物体的空间邻近关系,构建场景物体空间邻近图;进而基于该空间邻近图计算场景图像对的空间邻近度,完成图像空间关系匹配。试验表明不匹配图像间的空间邻近度一般为0,而匹配图像间的空间邻近度一般大于0,本文空间关系匹配涉及多个物体间的相互关系,具有更强的稳健性,其匹配效果明显优于对比试验中的其他方法,可以高效稳定地完成图像匹配任务。     

Data matching of building polygons at multiple map scales improved by contextual information and relaxation

The aim of matching spatial data at different map scales is to find corresponding objects at different levels of detail (LODs) that represent the same real-world phenomena. This is a prerequisite for integrating, evaluating and updating spatial data collected and maintained at various scales. However, matching spatial data is not straightforward due to the ambiguities caused by problems like many-to-many correspondence, non-systematic displacement and different LODs between data sets. This paper proposes an approach to matching areal objects (e.g. buildings) based on relaxation labeling techniques widely applied in pattern recognition and computer vision. The underlying idea is to utilize contextual information (quantified by compatibility coefficient) in an iterative process, where the ambiguities are reduced until a consistent matching is achieved. This paper describes (1) a domain-specific extension to previous relaxation schemes and (2) a new compatibility coefficient that exploits relative relationships between areal object pairs in spatial data. Our approach were validated through extensive experiments using building data sets at 1:10k and 1:50k as an example. Our contextual approach showed superior performance against a non-contextual approach in general and especially in ambiguous situations. The proposed approach can also be applied to matching other areal features and/or for a different scale range.     

利用地标空间关系约束的点实体匹配方法

针对传统的点实体匹配方法的不足,提出了利用地标空间关系约束的点实体匹配方法。该方法以参考点实体、候选匹配点实体与其邻域内的同名地标之间的空间关系为基础构造距离特征向量和方向特征向量,基于距离特征向量和方向特征向量计算点实体的相似度,进而确定点同名实体。实验结果表明,该方法在点同名实体空间位置偏差较大的情况下依然能够取得很好的效果,验证了该方法的有效性。     

利用概率松弛法的城市路网自动匹配

多源空间数据匹配是空间数据集成与互操作,变化检测与数据更新的重要前提。路网数据匹配在导航、智能交通和基于位置服务等领域具有重要的研究意义和实用价值。本文提出一种基于概率松弛方法的城市路网自动匹配方法,该方法首先通过路段间几何差异性估算候选路段的初始概率,然后根据邻接候选匹配路段的兼容性不断更新原概率矩阵直到收敛于某一极小值。最后基于收敛的概率矩阵计算各候选路段的结构相似性,并通过设定相应的规则选取和提炼1: 1, 1: M和M: N匹配对。实验选取中国武汉,瑞士苏黎世地区的OpenStreetMap数据与导航数据进行匹配算法的验证。结果表明：本文算法对非刚性偏差较大的路网数据能达到较高精度,不存在匹配方向性问题,且能够识别1: 0, 1: M和M: N匹配。     

顾及上下级空间关系相似性的道路网联动匹配方法

现有道路网匹配方法中,大多利用道路自身结点和弧段特征进行匹配,而较少注意道路邻域要素在道路网匹配中的重要定位参考作用,从而影响匹配效率和正确率的进一步提高。针对上述问题,本文提出了一种顾及上下级空间关系相似性的道路网联动匹配方法,即模仿人在读图时通过特征地物和空间关联寻找目标地物的思维过程,将匹配看作是一种特征目标寻找、信息关联传递的推理过程。首先,运用Stroke技术将复杂道路网进行等级划分。其次,通过道路骨架关联关系树构建道路网联动匹配模型。最后,选取高等级骨干道路作为起始特征对象,计算道路间的上下级空间关系相似性,逐级迭代使匹配信息在道路网联动匹配模型中传递,从而得到匹配结果。试验表明,本文算法缩小了待匹配数据的搜索范围,能够有效提高匹配正确率和效率,尤其在数据位移较大、存在非系统性几何位置偏差的情况下优势明显。     

一种热红外与可见光影像深度特征匹配方法

针对无人机热红外影像与光学卫星影像的匹配难题,提出一种基于异源地标数据集学习的深度局部特征匹配方法。首先,利用生成对抗网络学习热红外与可见光影像的灰度分布规律,并进一步合成用于特征提取模型训练的热红外影像地标数据集;然后,联合残差网络和注意力机制模型,从数据集中学习深度不变特征;最后,经过对不变特征的匹配、提纯等处理,获得像对的正确匹配点。试验测试了该方法的性能,并与KAZE、特征检测描述网络和深度局部特征模型进行了对比。结果表明,提出的方法对灰度、纹理、重叠率以及几何变化具有较强的适应性,且匹配效率较高,可为无人机视觉导航提供支撑。     

A Back‐Propagation Neural Network‐Based Approach for Multi‐Represented Feature Matching in Update Propagation

Spatial data infrastructures, which are characterized by multi‐represented datasets, are prevalent throughout the world. The multi‐represented datasets contain different representations for identical real‐world entities. Therefore, update propagation is useful and required for maintaining multi‐represented datasets. The key to update propagation is the detection of identical features in different datasets that represent corresponding real‐world entities and the detection of changes in updated datasets. Using polygon features of settlements as examples, this article addresses these key problems and proposes an approach for multi‐represented feature matching based on spatial similarity and a back‐propagation neural network (BPNN). Although this approach only utilizes the measures of distance, area, direction and length, it dynamically and objectively determines the weight of each measure through intelligent learning; in contrast, traditional approaches determine weight using expertise. Therefore, the weight may be variable in different data contexts but not for different levels of expertise. This approach can be applied not only to one‐to‐one matching but also to one‐to‐many and many‐to‐many matching. Experiments are designed using two different approaches and four datasets that encompass an area in China. The goals are to demonstrate the weight differences in different data contexts and to measure the performance of the BPNN‐based feature matching approach.     

Impacts of Positional Error on Spatial Regression Analysis: A Case Study of Address Locations in Syracuse, New York

Positional error is the error produced by the discrepancy between reference and recorded locations. In urban landscapes, locations typically are obtained from global positioning systems or geocoding software. Although these technologies have improved the locational accuracy of georeferenced data, they are not error free. This error affects results of any spatial statistical analysis performed with a georeferenced dataset. In this paper we discuss the properties of positional error in an address matching exercise and the allocation of point locations to census geography units. We focus on the error's spatial structure, and more particularly on impacts of error propagation in spatial regression analysis. For this purpose we use two geocoding sources, we briefly describe the magnitude and the nature of their discrepancies, and we evaluate the consequences that this type of locational error has on a spatial regression analysis of pediatric blood lead data for Syracuse, NY. Our findings include: (1) the confirmation of the recurrence of spatial clustering in positional error at various geographic resolutions; and, (2) the identification of a noticeable but not shockingly large impact from positional error propagation in spatial auto‐binomial regression analysis results for the dataset analyzed.     

A building polygonal object matching method based on minimum bounding rectangle combinatorial optimization and relaxation labeling

Volunteered geographic information (VGI) is an emerging phenomenon where anyone can create geographic information and share it with others. Compared with traditional authoritative geospatial data, it has several advantages, such as enriched data, instant updates, and low cost. The object matching method is widely used in VGI quality assessment and data updates. However, VGI matching faces certain challenges, such as the levels of detail that vary from object to object, the uneven distribution of data quality, and the automated matching requirement. To resolve these problems, this article proposes a new matching method that effectively combines the advantages of minimum bounding rectangle combinatorial optimization (MBRCO) and relaxation labeling. The proposed method (1) avoids setting the similarity threshold and weights and does not require training samples. This process is realized based on contextual information and optimization. (2) It overcomes the disadvantage that the MBRCO algorithm cannot distinguish adjacent buildings with similar shapes. Our approach is experimentally validated using two publicly available spatial datasets: OpenStreetMap and AutoNavi map. The experimental studies show that the proposed automatic matching method outperforms all the threshold-based MBRCO methods and achieves high accuracy with a precision of 97.8% and a recall of 99.2%.     

A spatial–temporal Hopfield neural network approach for super-resolution land cover mapping with multi-temporal different resolution remotely sensed images

The mixed pixel problem affects the extraction of land cover information from remotely sensed images. Super-resolution mapping (SRM) can produce land cover maps with a finer spatial resolution than the remotely sensed images, and reduce the mixed pixel problem to some extent. Traditional SRMs solely adopt a single coarse-resolution image as input. Uncertainty always exists in resultant fine-resolution land cover maps, due to the lack of information about detailed land cover spatial patterns. The development of remote sensing technology has enabled the storage of a great amount of fine spatial resolution remotely sensed images. These data can provide fine-resolution land cover spatial information and are promising in reducing the SRM uncertainty. This paper presents a spatial–temporal Hopfield neural network (STHNN) based SRM, by employing both a current coarse-resolution image and a previous fine-resolution land cover map as input. STHNN considers the spatial information, as well as the temporal information of sub-pixel pairs by distinguishing the unchanged, decreased and increased land cover fractions in each coarse-resolution pixel, and uses different rules in labeling these sub-pixels. The proposed STHNN method was tested using synthetic images with different class fraction errors and real Landsat images, by comparing with pixel-based classification method and several popular SRM methods including pixel-swapping algorithm, Hopfield neural network based method and sub-pixel land cover change mapping method. Results show that STHNN outperforms pixel-based classification method, pixel-swapping algorithm and Hopfield neural network based model in most cases. The weight parameters of different STHNN spatial constraints, temporal constraints and fraction constraint have important functions in the STHNN performance. The heterogeneity degree of the previous map and the fraction images errors affect the STHNN accuracy, and can be served as guidances of selecting the optimal STHNN weight parameters.     

卫星影像匹配的深度卷积神经网络方法

本文侧重于智能化摄影测量深度学习的第一个方面：深度卷积方法。传统的影像同名点对提取算法通常利用人工设计的特征描述符及其最短距离作为匹配准则进行匹配,其匹配结果易陷入局部极值,造成部分正确匹配点对的遗漏。针对这一问题,本文引入深度学习方法,设计了一种基于空间尺度卷积层的两通道深度卷积神经网络,采用其进行影像间的匹配模式学习,实现了基于深度卷积神经网络的卫星影像匹配。试验表明,在处理异源、多时相、多分辨率的卫星影像情况下,本文方法比传统匹配方法能提取到更为丰富的影像同名点对,且最终匹配提纯结果正确率优于90%。     

Machine learning innovations in address matching: A practical comparison of word2vec and CRFs

Record linkage is a frequent obstacle to unlocking the benefits of integrated (spatial) data sources. In the absence of unique identifiers to directly join records, practitioners often rely on text‐based approaches for resolving candidate pairs of records to a match. In geographic information science, spatial record linkage is a form of geocoding that pertains to the resolution of text‐based linkage between pairs of addresses into matches and non‐matches. These approaches link text‐based address sequences, integrating sources of data that would otherwise remain in isolation. While recent innovations in machine learning have been introduced in the wider record linkage literature, there is significant potential to apply machine learning to the address matching sub‐field of geographic information science. As a response, this paper introduces two recent developments in text‐based machine learning—conditional random fields and word2vec—that have not been applied to address matching, evaluating their comparative strengths and drawbacks.     

基于SIFT的宽基线立体影像密集匹配

提出基于对极几何和单应映射双重约束及SIFT特征的宽基线立体影像多阶段准密集匹配算法。算法包括三个阶段：①基于特征点的空间分布和信息熵选取一定数量的最优SIFT特征点集并进行最小二乘初始稀疏匹配及立体像对的基本矩阵和单应矩阵估计;②对于其余特征,利用同名核线倾斜角及SIFT特征的尺度信息对匹配窗口的仿射变换参数进行迭代优化及变形改正、提取仿射不变SIFT特征描述符,并基于双重约束信息及欧氏距离测度进行匹配;③考虑宽基线立体影像较低的特征提取重复率,对第②步左右影像中未能成功匹配的特征点,基于双向搜索策略,采用基于盒滤波加速计算的SSD测度在变形改正后的双重约束区域中进行匹配,并对匹配结果进行加权最小二乘拟合定位。实际的宽基线立体影像试验结果证明了算法的有效性,可为后续的三维重建提供较为可靠的密集或准密集匹配点。