城市居住区识别的Voronoi邻域方法与初步实践

Having been regarded as a new key point in expanding GIS application, geographical spatial recognition is the principal motive factor that has a great impact on the behaviors, therefore, practical computing is imperative. Urban community, be it a relative independent district, is surrounded by geographical entity group and its area occupied is about 50% in space. Taking the urban residential district as a breakthrough point, this topic exploits the recognition calculation method under two-dimensional digital map environment. Otherwise, in combination with the statistical incidents of criminal offence, quantity relationship between the feature of community spatial morphology and its existing behavior is initially tested and verified. In the sequence of spatial entity inputting, scope affirming, structure expressing and recognition resolving, taken as the core supported structure, the Voronoi neighbourhood contributes to the main research contents of the topic, including but not limited to: (1) Multi-characteristics classification method on geographical entity; (2) Voronoi k-order neighbourhood calculation model on urban community; (3) Hierarchical Voronoi data structure expression on urban community; (4) Recognition calculation on urban community; (5) Prototype software manufactures and its preliminary practice. Under the background of urban community recognition, the morphology variants is described, computed and analyzed practically in procedure, thus endeavoring to reveal criminal incentive accordingly. The achievement of this topic has a far-reaching influence on geographical spatial recognition calculation, GIS Voronoi theory and method.

地理空间认知是影响行为的主要动因，被认为是拓展新型GIS应用的关键所在，因此，开展实践性计算研究势在必行。居住区是地理实体群经围合所形成的相对独立区域，空间上约占城市面积的50%。课题以城市居住区为切入点，探索二维数字地图环境下的居住区识别计算方法，结合刑事犯罪统计数据，初步实践验证居住区空间形态特征与行为存在的定量关系。依据空间实体输入、范围确定、结构表达及识别求解顺序，以Voronoi邻域为核心支撑结构，主要研究内容包括：(1)地理实体多特征分类方法；(2)居住区Voronoi k阶邻域捕获模型；(3)居住区层次Voronoi图表达数据结构；(4)居住区识别计算；(5)原型软件研制与初步实践。以居住区识别为背景，阐明居住区形态变量的描述、计算与分析过程，从居住区空间形态角度初步实践盗窃犯罪行为的诱因，课题研究成果对发展地理空间认知计算、GIS Voronoi理论与方法具有重要价值。

居住区是城市的主要组成部分，课题以Voronoi邻域为基本工具，探索实现城市居住区的表达与识别方法。主要工作内容：(1) 顾及地形要素提出梯度Voronoi图，其在描述邻近与邻域方面具有更佳的性质，并实现梯度Voronoi图的构建算法；(2) 以地标作为参照进行居住区表达，从可视区域与几何形状出发，提出一种复合Voronoi几何特征的地标提取方法；(3) 以居住区道路为描述骨架，设计基于化简分割的示意性道路地图生成方法；(4) 通过航空图像提取居住区构筑物，在图像光流场、相对定向以及航空图像复原3个方向进行了深入研究。主要结论：(1) 地形表面特征可叠加反映至Voronoi图中，并由Voronoi图势力范围表达，所提出的梯度Voronoi图建立了场与对象之间的映射关系，其可进一步用于丰富GIS空间分析功能；(2) 附加地标与建筑物的示意性道路地图是城市居住区表达的一种有效方式，在居住区提取与可视化两个方面表现出优势；(3) 航空影像及无人机视频可用于提取城市居住区构筑物且具有较高精度，与此同时，观测区已有航片可有效地改善由于抖动等因素造成的影像模糊。课题取得成果：(1) 在《测绘学报》、《武汉大学学报(信息科科学版)》发表EI检索论文8篇，录用2篇；(2) 培养7名硕士研究生，其中1人获2018年度云南省优秀硕士学位论文。