基于随机补片和DeepLabV3+的建筑物遥感图像变化检测

为有效解决传统遥感图像变化检测预处理复杂的问题,提出一种基于随机补片和DeepLabV3+的建筑物遥感图像变化检测方法。以ResNet50特征提取网络为基础,创建DeepLabV3+语义分割网络,并在图像和标签中创建大小为224像素×224像素的随机补片作为网络输入,训练建筑物提取网络;修改建筑物提取网络输入层为6通道,通过矩阵运算将两期遥感图像转换为一幅6通道非RGB图像,利用转换后的非RGB图像进行网络训练并验证变化检测精度。实验1利用ENVI5.3软件,采用马氏距离法进行变化检测;实验2采用改进的U-Net网络和随机补片,完成网络训练和精度验证;实验3使用实验2的训练数据和验证数据,采用随机补片和DeepLabV3+网络进行变化检测网络训练及精度验证。实验结果表明,该方法实验1、实验2、实验3建筑物变化检测平均交并比分别为24.43%、83.14%、89.90%,边界轮廓匹配分数分别为61.47%,80.24%、96.51%。     

多源道路智能选取的本体知识推理方法EI北大核心CSCD

大数据时代道路数据来源日益增多,跨数据源的道路选取面临巨大挑战。本文针对数据语义不一致问题,提出一种基于本体知识推理的多源道路选取方法。首先,将1∶5万基本比例尺地形图道路数据作为基础案例,将四维图新导航电子地图和开放街道地图中的道路数据作为试验数据,基于stroke计算道路等级、长度、连通度、接近度、中介度特征项,提取特征项概念并构建本体;然后,从语义特征项和数值特征项两方面计算本体概念相似性,建立基础案例与试验数据间的关联关系;最后,基于本体和语义网规则语言定义本体通用、语义特征、数值特征三类选取规则,实现跨数据源道路选取的过程性知识推理。试验表明,本文方法可基于本体概念相似性度量消除语义差异,同时利用语义网规则语言进行知识推理,可实现多源道路数据向基本比例尺数据的智能选取。     

移动轨迹的尺度特征

移动轨迹是移动对象在地理空间环境中移动而留下的痕迹。移动轨迹数据是一类特殊的地理时空数据,它具有时间尺度、空间尺度和语义尺度特征。本文首先分析了轨迹数据的空间尺度特征与时间尺度特征,建立了轨迹空间尺度与时间尺度转换关系式;然后论述了轨迹的语义内涵和语义尺度特征,将轨迹语义分为移动对象语义、地理空间环境语义、采集设备语义、移动方式语义,并从地理空间环境语义的角度分析了轨迹的语义多尺度;最后探讨了轨迹语义尺度与时空尺度之间的一致性关系。     

Multi-level temporal autoregressive modelling of daily activity satisfaction using GPS-integrated activity diary data

In this research, we match web-based activity diary data with daily mobility information recorded by GPS trackers for a sample of 709 residents in a 7-day survey in Beijing in 2012 to investigate activity satisfaction. Given the complications arising from the irregular time intervals of GPS-integrated diary data and the associated complex dependency structure, a direct application of standard (spatial) panel data econometric approaches is inappropriate. This study develops a multi-level temporal autoregressive modelling approach to analyse such data, which conceptualises time as continuous and examines sequential correlations via a time or space-time weights matrix. Moreover, we manage to simultaneously model individual heterogeneity through the inclusion of individual random effects, which can be treated flexibly either as independent or dependent. Bayesian Markov chain Monte Carlo (MCMC) algorithms are developed for model implementation. Positive sequential correlations and individual heterogeneity effects are both found to be statistically significant. Geographical contextual characteristics of sites where activities take place are significantly associated with daily activity satisfaction, controlling for a range of situational characteristics and individual socio-demographic attributes. Apart from the conceivable urban planning and development implications of our study, we demonstrate a novel statistical methodology for analysing semantic GPS trajectory data in general.     

多模型协同下的城郊地区DEM构建方法研究

人类活动对地表的改造使其呈现形态多样化、不连续等特征,此时传统的DEM构建方法难以满足这些区域DEM精度要求。为此,本文提出了一种多模型协同构建DEM的思路,首先按照形态特征和语义信息对地形进行分类,然后对不同类型的地形区域,选择、设计适宜的方法分别进行DEM构建,最后将不同区域构建DEM结果融合拼接形成区域完整的DEM结果。本文选择江苏省南京市城市郊区某区域为实验区,以1:500比例尺地形图为基本数据源进行DEM构建实验。实验结果表明,与传统经典DEM构建方法相比,本文提出的多模型协同的DEM构建方法能够有效表达实验区域不同的地形特征,特别是对于人工改造的地形(如道路、边坡等区域),本文方法构建的DEM其形态精度优势显著;同时,基于验证点法的高程精度分析结果表明,本文方法构建DEM的高程精度亦优于传统DEM构建方法,特别是对于一些形态规则而高程信息相对稀少的区域,以边坡区域为例,经典DEM构建法平均误差均超过5 m,而本文构建结果平均误差为0.26 m,精度优势非常明显。研究表明本文提出的多模型协同的DEM构建方法适用于人类活动改造或显著影响的区域的DEM构建。     

主题学习和稀疏表示的MODIS图像超分辨率重建

针对MODIS图像分辨率受传感器限制和噪声干扰,且分辨率局限在一定水平等问题,提出一种采用主题学习和稀疏表示的MODIS图像超分辨率重建方法,该方法通过双边滤波将MODIS图像的平滑及纹理部分分离,并将纹理部分看成是由若干"文档"组成的训练样本;运用概率潜在语义分析提取"文档"的潜在语义特征,从而确定"文档"所属的"主题"。在此基础上,针对每个主题所对应的图像块,采用改进的K-SVD方法训练若干适用于不同主题的高低分辨率字典对,从而可以运用这些字典对,通过稀疏编码实现测试图像相应主题块的超分辨率重建。实验结果表明,重建图像在视觉效果和PSNR等指标上均优于传统方法。     

Extracting and analyzing semantic relatedness between cities using news articles

News articles capture a variety of topics about our society. They reflect not only the socioeconomic activities that happened in our physical world, but also some of the cultures, human interests, and public concerns that exist only in the perceptions of people. Cities are frequently mentioned in news articles, and two or more cities may co-occur in the same article. Such co-occurrence often suggests certain relatedness between the mentioned cities, and the relatedness may be under different topics depending on the contents of the news articles. We consider the relatedness under different topics as semantic relatedness. By reading news articles, one can grasp the general semantic relatedness between cities; yet, given hundreds of thousands of news articles, it is very difficult, if not impossible, for anyone to manually read them. This paper proposes a computational framework which can ‘read’ a large number of news articles and extract the semantic relatedness between cities. This framework is based on a natural language processing model and employs a machine learning process to identify the main topics of news articles. We describe the overall structure of this framework and its individual modules, and then apply it to an experimental dataset with more than 500,000 news articles covering the top 100 US cities spanning a 10-year period. We perform exploratory visualizations of the extracted semantic relatedness under different topics and over multiple years. We also analyze the impact of geographic distance on semantic relatedness and find varied distance decay effects. The proposed framework can be used to support large-scale content analysis in city network research.     

基于Solr的空间数据语义搜索研究

随着互联网技术的飞速发展,基于网络地图的空间数据搜索成为人们获取空间信息的重要手段。文章分析了当前地图搜索的不足和瓶颈,阐述了其在处理空间语义方面的缺陷,提出了一种基于Solr的空间数据语义搜索方案:将全文检索引擎Solr应用到空间数据搜索中;同时,引入自然语言处理和本体技术,实现基于自然语言查询的空间数据语义搜索。最后建立原型系统进行验证,证明了该方案的可行性和有效性。     

Exploration of spatiotemporal and semantic clusters of Twitter data using unsupervised neural networks

ABSTRACT

The investigation of human activity patterns from location-based social networks like Twitter is an established approach of how to infer relationships and latent information that characterize urban structures. Researchers from various disciplines have performed geospatial analysis on social media data despite the data’s high dimensionality, complexity and heterogeneity. However, user-generated datasets are of multi-scale nature, which results in limited applicability of commonly known geospatial analysis methods. Therefore in this paper, we propose a geographic, hierarchical self-organizing map (Geo-H-SOM) to analyze geospatial, temporal and semantic characteristics of georeferenced tweets. The results of our method, which we validate in a case study, demonstrate the ability to explore, abstract and cluster high-dimensional geospatial and semantic information from crowdsourced data.     

基于用户点击数据的细粒度图像识别方法概述

近年来,细粒度图像识别逐渐成为计算机视觉领域的研究热点.由于不同类别图像间的视觉差异小、语义鸿沟问题严重,传统的基于视觉特征的细粒度图像识别性能往往不尽人意.针对这些挑战,目前许多学者都在研究基于用户点击数据的图像识别.本文围绕点击数据在图像识别中数据预处理、特征提取和模型构建3大模块中的应用,总结了已有的基于点击数据的识别算法及最新的研究进展.