城市微环境与盗窃犯罪关系研究进展

城市微观社会环境和微观建成环境（即“城市微环境”）失序是影响盗窃犯罪时空格局和形成机理的两大关键因素,直接影响犯罪的机会、成本和手段,掌握微环境与盗窃的关系是CPTED、犯罪地理学的核心内容。本文聚焦国内外城市微环境与盗窃关系的研究进展,系统梳理了影响盗窃时空分异的城市微环境特征,总结了微环境的测度方式。总体上有4类问题待解决：① 当前利用手机大数据求算的环境人口指标主要为人口数量和流动性,少有从人口属性、社交规律等方面深层刻画微观社会环境。② 缺乏对微观建成、社会环境的系统性整合,易导致伪相关或不相关。③ 西方犯罪学理论的“本土化”实证研究不够充分,理论内涵和概念存在本土操作化困难,中国实证研究产生的“本土”知识对西方理论有补充性和挑战性,但还需“国际概念化”。④ 街景已被用于微观建成环境的虚拟测量,但随机选取测量地点的做法因未顾及微观社会环境的非均衡性而缺乏针对性。未来有4方面研究趋势：① 与公安部门紧密合作,结合敏感数据,刻画微观尺度人口属性、社交规律等特征,拓展“风险人口”的概念外延,准确反映微观社会环境,开展理论和政策导向的实证研究。② 耦合微观社会和建成环境,从社会、环境、行为等多维理论层面辨析两种微观环境的互嵌依据和耦合路径,诉诸“大”数据刻画“微”环境,探索微环境对盗窃的条件交互机制和尺度依赖性。③ 把握犯罪学理论内涵,兼顾中国城市社会情境,开发理论核心概念“本土”操作化新可能。④ 持续发展基于街景的远程化、虚拟化、自动化、智能化建成环境观测手段,注重针对性采样和系统性整合研究。     

美国海底地名标准化工作管窥

美国是开展海底地名标准化工作较早的国家之一,以其全球视野开展海底地名工作,在国际上掌握了较大的话语权。通过在美国地名委员会内设立海底地名咨询委员会,对海底地名的命名原则、通名选用等方面进行明确规定,不断完善纪念性命名管理制度等,美国形成了比较严密的地名标准化管理和技术体系。这些对我国的海洋管理和地名标准化工作都有较高的参考价值。     

Street networks in relation to landforms：Implications for fast-growing cities

In addition to socio-economic factors,major landforms may affect the city structure and urban form. Here we show that landforms have significant effects on the city shape and street patterns of the fast-growing Iranian cities of Dezful (a river) and Khorramabad (mountains and valleys),but no clear effects on the cities of Yazd and Nain. Also,where the street orientation is peaked,the Gibbs/Shannon entropy (a measure of dispersion or spread) is low,but increases as the distribution becomes more uniform because of landform constraints. The streets in the old inner parts of all the cities are,on average,shorter and denser (more streets per unit area) than the streets of the newer outer parts. The entropies of the outer parts are also greater than those of the inner parts,implying that the street-length distribution gradually becomes more dispersed or spread as the city expands. All these cities have been fast growing in the past decades,with the newer outer parts expanding rapidly. As shown here,the rapidly formed outer parts (with greater dispersion in street patterns) have significantly different textures from those of the older inner parts,indicating different functionality and growth processes. These quantitative methods for street-network analysis can be used worldwide,particularly for analysing the effects of landforms on city shape and texture.     

Exploring the relationship between street centrality and land use in Stockholm

This paper examines the relationship between different street centralities and land-use types in Stockholm. Major centrality measures of closeness, betweenness, and straightness are calculated at both global and local levels in both the primary and dual representations of the urban street network. Adaptive kernel density estimation is adopted to transform all unevenly distributed datasets to one continuous raster framework for further analysis. After computing statistical and spatial distribution of each centrality and land-use density map, we find that the density of each street centrality is highly correlated with one type of land use. Results imply that various centralities representing street properties from different aspects can capture the land development patterns of different land-use types by reflecting human activities, and are consequently important indicators to describe urban structure.     

基于街景影像多特征融合的广州市越秀区街道空间品质评估

全面准确地描述街景影像的多层次特征在基于街景影像对街道空间品质进行评估的研究中具有重要意义。以广州市越秀区为例,获取前后左右各视角街景影像中手工设计的特征（SURF特征、HOG特征）和基于深度学习的特征（语义特征）,基于单一特征和多特征融合采用支持向量机（Support Vector Machine, SVM）、随机森林（Random Forest, RF）训练各视角的评估模型。结果表明,以基于SVM建立的单特征模型为例,基于HOG特征（73.03%）、语义特征（72.28%）的模型平均精度优于SURF特征（56.00%）,基于SVM前后左右各视角模型的最优分类精度为82.8%（前）、81.7%（后）、76.6%（左）、76.6%（右）,而基于RF各视角模型的最优分类精度为82.8%（前）、85.0%（后）、78.1%（左）、70.3%（右）。前后视角的模型精度略优于左右视角。各视角最优模型均为多特征融合模型,最优模型平均分类精度和Kappa系数可达80.6%和0.62。利用街景影像评估街道空间品质时,各算法之间性能差异微弱,而特征选择及组合方式是提升精度的关键。越秀区街道空间品质存在明显空间分异,其西南部的街道空间品质亟待提升。本研究构建了基于街景影像多特征融合的大规模高精度街道空间品质测度方法,实现了对越秀区街道空间品质的评估,研究结果可为相关部门进行街道环境综合整治提供参考。     

基于人工神经网络的海底细粒土工程分类研究

海洋沉积物工程定名对于开展海洋工程建设具有重要作用,然而海底粉土和黏性土的定名受人为因素影响容易产生误差.使用人工神经网络的方法对黄河口埕岛海域284组细粒土数据进行了训练和学习,得到了只利用沉积物粒径质量分数进行定名的方法.结果表明,使用人工神经网络的方法能够有效地对沉积物进行工程定名.当网络含有5个输入层节点、9个...