OpenStreetMap道路数据质量的定量估计

The OpenStreetMap (OSM), as a free spatial data source based on crowdsourcing, has received wide applications. Due to the majority of OSM data is collected by non specialists and amateur geographers, one of the significant concerns is: how good is the quality of OSM data. Currently, extensive studies have been focused on assessing various quality measures (e.g. positional accuracy, attribute accuracy, logical consistency, completeness and so on) of OSM data through comparing with reference data. But, reference data is not always available, very few studies have discussed how to quantitatively estimate the quality of OSM data without reference data, which is the aim of our project. This project pays attention to two essential issues: first, which indicators can be used as proxies for quantitatively estimating various quality measures of OSM road data; second, how to quantitatively establish the relationship between proxy indicators and various quality measures of OSM data. For the first issue, this project analyzes the correlation between a number of factors (including spatial data factor, regional geography factor, social statistical factor and historical editing factor) and various quality measures of OSM road data, and proposes proxy indicators for each quality measure. For the second issue, this project proposes to use regression approaches and machine learning approaches to quantitatively establish the relationship between proxy indicators and various quality measures of OSM road data. The above proxy indicators and approaches will also been applied to quantitatively estimate the quality of updated road data in OSM compared to authorized data, for the purpose of map updating. More importantly, the above proxy indicators and approaches may also be applied to quantitatively estimate OSM data quality of other geographic features (e.g. buildings, railways and land uses) and to update them in authorized data, in order to meet the need of acquiring the latest data for a specific area in a short period of time, which has great significance on map navigation, national geographic state monitoring, disaster prevention and reduction and so on.

OpenStreetMap（OSM）是最典型的一种自发地理信息，应用十分广泛。然而OSM数据一般由非专业人员提供，存在数据质量问题。本项目针对当前OSM数据质量评价方法依赖参考数据，无法定量估计未知参考数据的OSM数据质量，研究如何确定OSM道路数据质量定量估计的代理指标和如何构建代理指标与OSM道路数据质量的关联模型等关键科学问题。通过分析OSM道路数据质量与空间数据因子、区域地理因子、社会统计因子和历史编辑因子的相关性，提出OSM道路数据质量估计的代理指标；通过构建代理指标与OSM道路数据质量的回归模型和机器学习模型，实现定量估计未知参考数据的OSM道路数据质量；还以地图更新应用为例，提出通过定量估计OSM增量道路数据，增强OSM数据的实用性。本项目的研究思路可拓展用于建筑、铁路和土地利用等OSM其它地理要素的质量估计及快速更新，对满足用户在较短时间内获得特定地区最新数据具有积极意义。

OpenStreetMap（OSM）是由全球志愿者在线提供的众源地理数据，应用广泛。虽然OSM数据具有免费获取、全球覆盖和现势性高等优点，但是存在数据质量问题。针对现有的OSM数据质量评价方法普遍依赖参考数据的不足，本项目提出（1）当参考数据未知时，定量估计OSM数据质量的方法。提出基于街区面积、街区周长和街区密度指标定量估计OSM道路数据质量的方法；提出基于建筑密度指标定量估计OSM建筑数据质量的方法；提出以多样性作为现有数据质量标准的补充，估计OSM土地覆盖/土地利用数据质量的方法；提出一种自动构建OSM与参考数据土地覆盖/土地利用类别对应关系的方法。（2）基于所提出的指标与方法，以省/市级行政区划、城市建成区和公里格网等为分析单元，揭示了我国OSM道路数据中缺失的道路类型；揭示了我国OSM建筑和土地覆盖/土地利用数据的空间分布格局及其影响因素。（3）拓展了OSM道路数据支持下的城市规划和可持续发展应用。提出基于OSM道路数据的城市原点选址的几何、拓扑和语义方法；提出基于OSM数据的城市形态分析方法，并揭示了我国2000多个城市建成区的街道网形态及其空间格局；提出OSM道路数据等多源数据支持下的农村可达指数分析方法，并揭示了非洲55个国家难以享受道路交通服务的农村人口。本项目提出的指标与方法不仅弥补了现有空间数据质量评价方法依赖参考数据的不足，还能服务于获取与分析全球范围的地理信息数据和支撑联合国2030年可持续发展目标等应用领域。研究成果已发表SCI/SSCI论文11篇，包括发表在地理信息科学领域国际顶级学术期刊IJGIS的论文2篇；申请国家发明专利2项；获中国测绘学会青年优秀论文二等奖和湖北省社会科学优秀成果三等奖；项目成员参加国内外学术会议并作口头报告20余次；指导硕士生13人，1人获研究生国家奖学金和优秀硕士学位论文；撰写咨询建议报告2项，获武汉市市长批示1项。