Building-Level Change Detection from Large-Scale Historical Vector Data by Using Direct and a Three-Tier Post-classification Comparison
Building-Level Change Detection from Large-Scale Historical Vector Data by Using Direct and a Three-Tier Post-classification Comparison
复制标题
使用直接和三层后分类比较从大规模历史矢量数据检测建筑物级别变化
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Dongmei Chen
中科院分区：
文献类型：
--
作者：
M. Hussain;Dongmei Chen
城市地区建筑层面的历史变化信息对于政策和资源管理至关重要，特别是在人口密集、建筑建设速度快的国家。在本文中，我们提出了一种使用大规模历史向量和基于地址的数据的多级建筑变更检测框架。这种方法与传统方法完全不同，传统方法纯粹使用遥感图像，并且在识别功能特征方面往往受到限制。以英国地形测量局（OS）的MasterMap作为大比例尺矢量数据的例子。提取两年来的建筑物特征并进行比较，以识别修改过的、拆除过的和未更改过的建筑。为了量化建筑物的功能变化，使用了早期开发的分类方法，通过提取建筑物的图表和空间属性并从基于地址的数据链接上下文信息。英国曼彻斯特的案例研究表明，所提出的方法可以成功地识别多个层面的建筑变化。这里提出的变化检测框架密切关注如何使用大规模和现有的数据源来创建历史土地利用数据库。此外，该框架计算稳健，适用于其他领域，而不会失去英国境内外的完整性，因为英国境内和境外都有大规模结构化数据集。
Historical change information at the building level in urban areas is crucial for policy and resource management, especially in countries with densely population and quick building construction. In this paper we present a multi-level building change detection framework using large-scale historical vector and address-based data. This approach is fundamentally different to the traditionally ones which purely use remotely sensed images and are often limited in identifying functional characteristics. Ordnance Survey’s (OS) MasterMap in the UK has been taken as an example of the large-scale vector data. The buildings features are extracted for two years and are compared to identify modified, demolished, and un-changed ones. To quantify buildings’ functional changes, an earlier developed classification methodology was used by extracting cartomteric and spatial properties of buildings and linking contextual information from address-based data. The case study in Manchester, UK shows that the proposed approach can successfully identify building changes at multiple levels. The change detection framework presented here closely addresses how to use large-scale and existing data sources to create a historical land use database. Moreover, this framework is computationally robust and is applicable to other areas without losing its integrity within and outside the UK, where large-scale structured data sets are available.