On the spatial distribution of buildings for map generalization

Information on spatial distribution of buildings must be explored as part of the process of map generalization. A new approach is proposed in this article, which combines building classification and clustering to enable the detection of class differences within a pattern, as well as patterns within a class. To do this, an analysis of existing parameters describing building characteristics is performed via principal component analysis (PCA), and four major parameters (i.e. convex hull area, IPQ compactness, number of edges, and smallest minimum bounding rectangle orientation) are selected for further classification based on similarities between building characteristics. A building clustering method based on minimum spanning tree (MST) considering rivers and roads is then applied. Theory and experiments show that use of a relative neighbor graph (RNG) is more effective in detecting linear building patterns than either a nearest neighbor graph (NNG), an MST, or a Gabriel graph (GssG). Building classification and clustering are therefore conducted separately using experimental data extracted from OpenStreetMap (OSM), and linear patterns are then recognized within resultant clusters. Experimental results show that the approach proposed in this article is both reasonable and efficient for mining information on the spatial distribution of buildings for map generalization.     

多尺度SLIC-GMRF与FCNSVM联合的高分影像建筑物提取

遥感影像建筑物提取具有重要的应用价值。然而,高分辨率遥感影像中细节信息繁多、特征复杂,增加了建筑物提取难度。针对这一问题,本文提出一种基于多尺度SLIC-GMRF和FCNSVM的建筑物提取方法,一定程度上提高了高分辨率遥感影像建筑物提取能力。首先,利用多尺度SLIC-GMRF分割算法确定初始建筑物区域,然后,充分利用FCN神经网络在语义分割中的优势抽取建筑物特征,最后,结合提取出的建筑物特征训练SVM分类器细化建筑物提取结果,通过3种控制实验,两种对比方法得出以下结论:SLIC分割算法影响初始分割结果;SVM分类器影响建筑物细部提取;FCN特征影响SVM分类器性能。对于特征清晰、遮挡干扰较少的研究区,本文方法能够较好提取影像中的建筑物,查准率、查全率、质量指标均优于对比方法,对建筑物复杂分布的研究区同样能够取得较好的提取效果。     

一种建筑群同质二元组直线模式的识别方法

建筑群空间分布模式识别对制图综合、多尺度表达及空间数据挖掘具有重要意义。针对建筑群中以建筑物组合结构为单元的直线模式识别问题,提出一种建筑群同质二元组直线模式的识别方法。首先分析研究同质二元组直线模式的认知特征和定义;然后利用Delaunay三角网构建建筑群邻近关系,以建筑物邻近性、尺寸和方向相似性约束进行聚类,考虑邻近程度的异质性、直线性识别直线模式,根据直线模式识别结果识别同质二元组结构;最后逐步识别由具有相似特征且呈等间隔直线排列的同质二元组构成的直线模式。实验表明所提方法能有效识别出同质二元组直线模式,其识别结果符合人类认知。     

Analyzing the effects of scale and land use pattern metrics on land use database generalization indices

The generalization index system is one of the critical issues for computer-aided land use database generalization. This paper studies the scale and land use pattern effects on land use database generalization indices and estimates the thresholds of these indices based on a typical land use database sample. The index system of land use database generalization is discussed and constructed from macro and micro perspectives. Six land use pattern metrics, namely, land use diversity index, land use dominance index, land use homogeneity index, land use fragmentation index, the index of land use type dominance, and the index of land use type fragmentation, are designed to characterize land use patterns and are introduced into the analysis of land use pattern effect on land use database indices. The analysis framework of the scale and land use pattern effects on the land use database indices are proposed by employing statistical techniques. Based on the land use database samples at multiple spatial scales collected in various land use regions across China, the study generates rules for both scale and land use pattern effects on the indices, including map area proportion of land use types, total map load, parcel map load, and minimum parcel area. The thresholds of these indices in land use database generalization are produced at the scales of 1:50,000, 1:100,000, 1:250,000, and 1:500,000. An experimental generalization at county level demonstrates how to determine the generalization index values considering scale and land use pattern, and how to evaluate the generalization results using our macro indices.     

基于U型卷积神经网络的航空影像建筑物检测

经典的卷积神经网络结构在前向传播过程中分辨率不断下降,导致仅采用末层特征时难以实现建筑物边缘的精确分割,进而限制目标检测精度。针对该问题,提出一种基于U型卷积网络的建筑物检测方法。首先借鉴在图像分割领域中性能出色的神经网络模型U-Net的建模思想,采用对称式的网络结构融合深度网络中的高维和低维特征以恢复高保真边界;其次考虑到经典U-Net对位于特征金字塔顶层的模型参数优化程度相对不足,通过在顶层和底层两个不同尺度输出预测结果进行双重约束,进一步提升了建筑物检测精度。在覆盖范围达30 km²、建筑物目标28 000余个的航空影像数据集上的试验结果表明,本文方法的检测结果在IoU和Kappa两项关键评价指标的均值上分别达到83.7%和89.5%,优于经典U-Net模型,显著优于经典全卷积网络模型和基于人工设计特征的AdaBoost模型。     

基于北京二号影像辅助nDSM的建筑物自动提取

针对建筑物提取方法缺乏泛化性的问题,本文提出了将nDSM、北京二号影像、NDVI、BAI的七通道图像相结合作为数据源的提取方法。采用随机森林、梯度提升机、支持向量机、BP神经网络分类器对建筑物进行提取获取最佳分类器模型,并运用二值化与开闭运算,以建筑物面积与最小外接矩形面积的比值为阈值,对建筑物分别进行最小外接矩形、DP算法拟合,优化建筑物提取结果。试验结果表明,梯度提升机(GBDT)较其他分类模型在不同场景下综合效果较好,F-score精度更高。     

顾及功能语义特征的建筑物空间分布模式识别方法

本文从空间-语义双重约束角度,提出一种顾及空间邻近和功能语义相似的建筑物空间分布模式识别方法。首先,基于建筑物的空间位置邻近性(即建筑物间的最小距离)约束进行聚类,获得建筑物的空间分布模式和建筑物间的空间邻近关系;然后,根据建筑物的功能语义相似性约束进行分割,获得建筑物的初步聚类结果;最后,考虑簇内相似性与簇间差异性进行整体优化,获得最终聚类结果。试验验证表明,本文方法比现有方法能够更有效地识别空间邻近与功能语义一致的建筑物群,服务于智慧城市建设中对建筑物进行语义层次综合和对城市结构进行深入研究的需求。     

直线排列建筑物群渐进式典型化方法

地图综合中,建筑物群的排列结构是需要重点考虑的因素。当不同排列的子建筑物群之间存在空间图形冲突时,这些建筑物群的综合就显得更为复杂。直线排列建筑物群的综合在大比例尺地形图上以典型化操作为主。本文提出一种相互之间存在潜在空间图形冲突的多个直线排列建筑物子群的渐进式典型化方法,渐进式地处理多个直线排列建筑物子群之间的空间图形冲突,保留建筑物群重要的直线排列结构;以建筑物表达的视觉图形约束为限制条件,自动确定典型化后的建筑物位置、形状、大小和方位。本文还研究了基于建筑物群空间邻近图的直线排列建筑物子群的自动识别方法,分析了这些直线排列之间的邻近关系和相交关系。最后,以1：5000地图上的建筑物群综合为1：25 000为试验对象,验证了所提出算法的可用性和有效性。     

A typification method for linear pattern in urban building generalisation

This paper presents a typification method for linear pattern in urban building generalization. The proposed method includes two processes. Firstly, structural knowledge in terms of linear pattern is detected using a two-step algorithm taking the advantages of Gestalt visual perception, computational geometry and graph theory. Spatial neighbourhood is captured using interpolated constrained Delaunay triangulation and the resulting proximity graph is pruned to be heterogeneous to get acceptable linear patterns with regard to Gestalt visual perception. Then, a typification strategy is proposed, in which typification is regarded as a progressive and iterative process consisting of elimination, exaggeration and displacement. The typification strategy iteratively executes eliminating the building with minimum overall effect, exaggerating remaining buildings considering key location and spatial characteristics and displacing them to preserve the linear pattern until elimination quantity is satisfied. Experiments show that this proposed strategy is effective and linear patterns are guaranteed with correctness and completeness.     

多尺度变换中建筑群空间相似度计算模型

针对建筑群自动综合结果质量评价问题,该文提出一种建筑群多尺度表达中空间相似度计算模型。从地图综合质量评价的角度,立足地图综合的3个约束,即图形约束、拓扑约束、结构约束,分别计算了图形相似度、拓扑相似度、结构相似度,并对以上各相似度作了适当的权重分配,计算建筑群目标的整体相似度。最后,通过实验验证了本文相似度计算方法的正确性、可行性。     

Categorical Database Generalization

This paper focuses on the issues of categorical database generalization and emphasizes the roles of supporting data model, integrated datamodel, spatial analysis and semantic analysis in database generalization. The framework contents of categorical database generalization transformation are defined. This paper presents an integrated spatial supporting data structure, a semantic supporting model and similarity model for the categorical database generalization. The concept of transformation unit is proposed in generalization.     

A spatiotemporal mining framework for abnormal association patterns in marine environments with a time series of remote sensing images

A spatiotemporal mining framework is a novel tool for the analysis of marine association patterns using multiple remote sensing images. From data pretreatment, to algorithm design, to association rule mining and pattern visualization, this paper outlines a spatiotemporal mining framework for abnormal association patterns in marine environments, including pixel-based and object-based mining models. Within this framework, some key issues are also addressed. In the data pretreatment phase, we propose an algorithm for extracting abnormal objects or pixels over marine surfaces, and construct a mining transaction table with object-based and pixel-based strategies. In the mining algorithm phase, a recursion method to construct a direct association pattern tree is addressed with an asymmetric mutual information table, and a recursive mining algorithm to find frequent items. In the knowledge visualization phase, a “Dimension–Attributes” visualization framework is used to display spatiotemporal association patterns. Finally, spatiotemporal association patterns for marine environmental parameters in the Pacific Ocean are identified, and the results prove the effectiveness and the efficiency of the proposed mining framework.     

全局局部细节感知条件随机场的高分辨率遥感影像建筑物提取

条件随机场模型由于其较强的上下文信息建模能力,被广泛应用于建筑物提取任务中。然而,面对高分辨率遥感影像丰富的地物信息,基于条件随机场的提取方法存在建筑物边界模糊的问题。本文提出了一种全局局部细节感知条件随机场框架,该框架提出全局局部一体化D-LinkNet,在有效利用多尺度建筑物信息的同时保留局部结构信息,解决了传统条件随机场一元势能丢失边界信息的问题。同时,该框架融合分割先验以缓解建筑物类内光谱差异较大的影响,利用更大尺度的上下文信息来精确提取建筑物,并引入局部类别标记代价从而保持细节信息以获取清晰的建筑物边界。实验结果表明,该框架在WHU卫星和航空数据集上的精度评价指标均优于其他对比方法,其IoU分别达89.82%和91.72%,对于复杂场景下的建筑物信息能够获得较好的提取效果。     

基于物方空间几何约束最小二乘匹配的建筑物半自动提取方法

提出了一种从数字航空立体像对半自动提取建筑物的方法。操作员通过人机界面选择房屋的种类并输入初始位置，然后经过边缘检测、直线段提取和据此房屋几何模型的线段自动编组等处理得到各房屋角点的初始位置，最后，为了获得房屋的精确定位和符合物方严格几何约束的结果，基于物方空间几何约束的最小二乘匹配平差模型用于求取房屋直线边缘和物方几何模型的最优匹配，试验表明，该方法能提高建筑物的提取精度并可为半自动建筑物提取提供一个灵活的框架。     

基于图匹配的城市建筑群典型字母型分布的识别

建筑群空间分布结构对于制图综合和多尺度表达等具有重要意义。结合国内外对该问题的研究,从结构模式识别的角度提出了基于图匹配算法的建筑群典型字母型分布模式的识别方法。首先统计和提取感兴趣的字母型分布模式,选定基元,选取合理的属性信息参数和结构信息参数,利用属性关系图形式语言描述模式,构建模板库。然后对建筑群抽象和降维,将其转化为基于属性关系图表达的场模型。最后通过Ullman图匹配算法求解非精确子图同构问题,从而识别建筑群中的典型字母型分布模式子集。实验表明该方法能够有效识别典型字母型建筑群分布,并为地图综合提供了新思路。     

Extracting urban water bodies from high-resolution radar images: Measuring the urban surface morphology to control for radar’s double-bounce effect

Satellite remote sensing is an effective method for extracting water bodies on a large scale. Radar imagery, such as synthetic aperture radar (SAR) imagery, can penetrate clouds and provide opportunities for water body identification when in situ observations are difficult to obtain because of severe weather conditions. However, when using SAR images in urban areas to extract water bodies, the radar’s double-bounce effect results in complicated backscatter patterns of water near urban features such as buildings due to the side-looking properties of SAR sensors and the vertical urban structures. Therefore, the objective of this study is to propose a reliable urban water extraction framework for SAR images that integrates urban surface morphological features for controlling radar’s multiple bounces. Statistical (logistic regression) and machine-learning (random forest) models were used to explore how radar’s double-bounce effect influences the prediction performance of urban water extraction. Our findings indicate that when extracting urban water bodies, urban water’s backscatter values could be significantly interfered by the neighboring building density above a threshold height that contributes to radar’s multiple bounces. Without model calibration, our framework incorporating urban surface morphology demonstrates high prediction ability with an Area Under the Curve (AUC) of 0.914 and with 97.0% of urban water cells correctly identified by testing in another city sharing similar urban forms. In summary, our study provides a better understanding of the role of the urban surface morphology in the double-bounce effect in SAR images, specifically for differentiating urban water and land, thereby improving the accuracy of urban water extraction and enhancing the feasibility of further applications of SAR imagery under complex urban landscapes.     

一种结合地理知识的遥感影像目标实体关联方法北大核心CSCD

遥感影像目标识别技术已广泛应用于目标动态监测与定位等领域。但影像目标识别的结果缺乏与目标属性信息的链接,导致分析人员只能依据影像特征进行分析,难以进行更复杂的目标数据关联分析与挖掘。针对遥感影像目标识别语义属性信息缺失的问题,本文利用知识图谱相关技术将影像判别的目标信息与知识语义网链接。首先,提出了一种遥感影像目标知识图谱构建框架;其次,针对遥感影像目标不同的数据类型,构建遥感影像目标知识抽取模型,提出了基于相似度目标实体识别和预定义模式的关系抽取方法;然后,基于多特征Logistic模型的影像目标实体链接方法,实现了遥感影像目标实体与百科知识库的知识关联;最后,针对预定试验区域进行试验,验证了本文方法的可行性。     

顾及尺度效应和景观格局的土地利用数据综合指标研究

从尺度效应和土地利用案观格局影响两个层面研究土地利用数据综合的宏观和微观综合指标,设计土地利用数据综合的指标体系和指标定量化描述模型.在全国各个土地利用分区采集不同尺度的具有典型代表性的土地利用数据样本集,通过空间统计分析等方法,得到土地利用数据综合中的宏观控制性指标和微观图形综合指标的尺度变化规律及其与土地利用格局的关系.以一个县为试验区,验证其合理性.

Abstract：

Indices for land use data generalization are critical for generating multiscale land use maps and databases. Previous research suffers from two major setbacks. Firstly, determination of threshold values for multi-scale land use data generalization in a large area, such as a nation, remains subjective. A second problem is general ignorance of landscape pattern. This paper studies the index system of land use data generalization from both scale and landscape pattern perspective. We discuss the macro and micro indices of land use data generalization with consideration of spatial scales and landscape pattern. To quantitatively relate the indices and scale and landscape pattern metrics, land use data samples have been collected at multiple spatial scales in various land use regions across China. Based on statistic analysis, we then generate both macro and micro control rules for land use data generalization at various spatial scales and land use patterns. Finally, we prove the proposed method and achieved results to be effective and reasonable with sample data at county level.     

车载激光扫描点云中建筑物边界的快速提取

以车载激光扫描点云数据为研究对象,提出一种由粗到细且快速获取点云中建筑物3维位置边界的方法。首先,通过分析格网内部点云的空间分布特征(平面距离、高程差异和点密集程度等)确定激光扫描点的权值,采用距离加权倒数IDW(Inverse Distance Weighted)内插方法生成车载激光扫描点云的特征图像。然后,采用阈值分割、轮廓提取与跟踪等手段提取特征图像中的建筑物目标的粗糙边界。最后,对粗糙边界内部的建筑物目标点云进行平面分割,提取建筑物的立面特征并构建立面不规则三角网TIN(Triangulated Irregular Network),并在建筑物先验框架知识条件下自动提取建筑物的精确3维位置边界。     

京津冀城市群1990—2015年城区时空扩展监测与分析

城区作为城镇化表征的重要指标,在城镇化时空扩展分析中有着重要的意义。然而,目前城区边界存在相关概念混杂导致对其认识缺乏统一、划定标准和数据来源缺乏一致性、遥感监测所用影像分辨率较低、以建设用地（不透水面）替代城区等问题。针对上述问题,参考住房和城乡建设部的建成区概念和城区地域分布范围,充分挖掘高分辨率遥感影像展现的城市景观和形态等空间可视化特征,从城市实际建设完成情况出发,本文提出了基于高分辨率影像和地理信息资料辅助的城区半自动化提取方法。获取了京津冀城市群153个县级以上城市的1990、2002、2015年的遥感影像、多元地理信息参考资料,开展了该区域的城区边界提取,并从时空扩展过程、城区空间形态及结构变化、城区扩展协调性、城区扩展占用土地类型4个方面对京津冀城市群1/4世纪的城镇化过程进行了统计分析。提出的技术方法和监测分析结果对于城镇化发展监测和城市规划实施评估具有重要的参考价值。