面向多源地理空间数据的知识图谱构建

知识图谱广泛应用于人工智能领域,基于此融合多源地理空间数据并表示地理事物的语义和时空信息,实现“数据—知识”的转换成为人们关注的热点。但现有通用知识图谱的空间知识覆盖度低且存在错误,同时基于维基百科构建的地理知识图谱存在空间关系、中文属性和坐标信息等属性缺失问题。因此本文以地理空间数据和百度百科数据的特征分析为基础,提出了以地理空间数据提取地理实体为主,百度百科补充属性信息为辅的知识图谱构建方式。① 基于GeoSparql设计模式层的地理实体、要素、几何形状和空间关系的逻辑关系;② 通过地理实体提取、实体链接和属性信息填充,在数据层实现空间知识融合;③ 结合关系型数据库和图数据库,设计空间知识存储方式;④ 在实体和关系2个方面定量分析知识图谱的构建规模。结果表明,本文构建的知识图谱中地理实体覆盖度和链接百科成功率相对较高,扩充了地理实体的概念描述信息,并将地理坐标的覆盖率提高到100%,对地理数据到地理知识的拓展具有重要意义。     

Gorgan Bay:a microcosm for study on macrobenthos species-environment relationships in the southeastern Caspian Sea

The relationship between spatial patterns of macrobenthos community characteristics and environmental conditions(salinity, temperature, dissolved oxygen, organic matter content, sand, silt and clay) was investigated throughout the Gorgan Bay in June 2010. Principal components analysis(PCA) based on environmental data separated eastern and western stations. The maximum(4500 ind./m²) and minimum(411 ind./m²) densities were observed at Stas 1 and 6, respectively. Polychaeta was the major group and Streblospio gynobranchiata was dominant species in the bay. According to Distance Based Linear Models results, macrofaunal total density was correlated with silt percentage and salinity and these two factors explaining 64% of the variability while macrofaunal community structure just correlated with salinity(22% total variation). In general, western part of the bay showed the highest number of species and biodiversity while, the highest density was found at Sta. 1 and in the middle part of the bay. Furthermore, relationship between diversity indices and macrobenthic species with measured factors is also discussed. Our results confirm the effect of salinity as an important factor on distribution of macrobenthic fauna in south Caspian brackish waters.     

资源三号多光谱影像谱段间相对内参关系标定及高精度配准

针对资源三号多光谱相机,提出利用同时获取的各谱段影像间高精度提取的同名像点建立多光谱相机各谱段对应分片CCD之间的位置模型,求解它们的相对内参关系参数,并应用此模型实现了谱段间影像的高精度谱段配准。试验结果表明该方法能够实现资源三号卫星多光谱影像优于0.3像素精度的谱段间自动配准。     

利用城市骨架线网的道路和居民地联动匹配方法

同名道路或者居民地数据间由于数据一致性程度不高,往往存在较大的几何位置偏差,当对道路或者居民地单独进行匹配时,不利于提高匹配正确率和效率。针对这一问题,本文提出了一种利用城市骨架线网的道路和居民地联动匹配方法。联动匹配即模仿人在读图时通过特征地物和空间关联寻找目标地物的思维过程,将匹配看作是一种特征目标寻找、信息关联传递的推理过程。首先,利用约束Delaunay三角网对地图构建城市骨架线网。然后,通过道路、骨架线、骨架线网眼和居民地之间的拓扑关系建立道路和居民地之间的匹配传递模型。最后,根据该传递模型实现通过道路匹配带动居民地匹配或者通过居民地匹配带动道路匹配的联动匹配。该方法优势在于只要有一种要素的数据一致性比较好,就能带动另一种要素取得很好的匹配效果,同时符合人类进行匹配时的认知过程。     

平面拓扑模型的多边形矢量更新算法

为了提高土地利用变更图斑的更新效率,文章提出了一种平面拓扑图模型下的多边形矢量数据更新算法:采用平面拓扑图的数据结构,实现多边形几何对象间的线段求交计算,解决多边形间在重点和重边情况下的几何裁剪问题;通过对底图几何对象与变更几何对象之间拓扑关系的判别,不仅能够检测出矢量底图数据中发生变化的几何信息,而且可以减少几何运算的次数,提升矢量数据的更新效率。实验结果表明该算法具有良好的适用性及较高的更新效率。     

1∶10000 DLG中高压线与铁塔点一致性自动处理方法的研究

为提高处理1∶10 000 DLG中高压线与铁塔点不一致性问题的工作效率,提出了两种方法:基于网络分析的高压线与铁塔点一致性自动处理;基于空间关系的高压线与铁塔点一致性自动处理。以某一测区内2 248条高压线与10 379个铁塔点作为实验数据,实验结果表明,本文提出的方法与人工处理相比在保证质量的前提下还极大地提高了工作效率。     

新型耦合数据驱动模型在降雨径流模拟中的应用研究

为解决传统数据驱动模型的不足,使其能实现降雨径流过程高精度连续模拟,提出新型耦合数据驱动模型——PEK,即:基于偏互信息的输入变量选择、基于新型集成神经网络的出流量预测和基于K最近邻模型的出流量误差预测。PEK模型具有以下特点:(1)提出了基于分离式选择策略和滑窗累积雨量的模型候选输入向量,并与基于偏互信息的输入变量选择方法联合使用,提高了输入信息的充分性和无冗余性,对建立精度高、泛化能力强的高质量模型意义重大;(2)提出了新型集成神经网络——EBPNN及其率定方法。联合使用NSGA-II多目标优化算法和早停止Levenberg-Marquardt算法,通过一次优化过程同时确定全局最优个体网络个数、各个体网络拓扑结构和网络参数。个体网络权重由基于AIC信息准则的权重优选方法确定。EBPNN在模拟精度和网络复杂度间取得了良好折衷,精度高、泛化能力强、率定结果客观;(3)PEK模型能够进行多步外推预报,实现了非实时校正模式下的高精度连续模拟,增长了预见期;(4)PEK模型不需要进行流域状态变量的计算,仅需初始出流量就可进行出流量的连续模拟。在呈村流域应用PEK和CLS两个数据驱动模型进行次洪降雨径流模拟及精度比较。结果表明PEK模型使用简便,模拟精度高于CLS模型,实现了多步外推的高精度连续模拟,增长了数据驱动模型的预见期。     

Assessing urban rainfall-runoff response to stormwater management extent

We report an empirical analysis of the hydrologic response of three small, highly impervious urban watersheds to pulse rainfall events, to assess how traditional stormwater management (SWM) alters urban hydrographs. The watersheds vary in SWM coverage from 3% to 61% and in impervious cover from 45% to 67%. By selecting a set of storm events that involved a single rainfall pulse with >96% of total precipitation delivered in 60 min, we reduced the effect of differences between storms on hydrograph response to isolate characteristic responses attributable to watershed properties. Watershed-average radar rainfall data were used to generate local storm hyetographs for each event in each watershed, thus compensating for the extreme spatial and temporal heterogeneity of short-duration, intense rainfall events. By normalizing discharge values to the discharge peak and centring each hydrograph on the time of peak we were able to visualize the envelope of hydrographs for each group and to generate representative composite hydrographs for comparison across the three watersheds. Despite dramatic differences in the fraction of watershed area draining to SWM features across these three headwater tributaries, we did not find strong evidence that SWM causes significant attenuation of the hydrograph peak. Hydrograph response for the three watersheds is remarkably uniform despite contrasts in SWM, impervious cover and spatial patterns of land cover type. The primary difference in hydrograph response is observed on the recession limb of the hydrograph, and that change appears to be associated with higher storm-total runoff in the watersheds with more area draining to SWM. Our findings contribute more evidence to the work of previous authors suggesting that SWM is less effective at attenuating urban hydrographs than is commonly assumed. Our findings also are consistent with previous work concluding that percent impervious cover may have greater influence on runoff volume than percent SWM coverage.     

A comparative study of nonlinear Markov chain models for conditional simulation of multinomial classes from regular samples

Simulating fields of categorical geospatial variables from samples is crucial for many purposes, such as spatial uncertainty assessment of natural resources distributions. However, effectively simulating complex categorical variables (i.e., multinomial classes) is difficult because of their nonlinearity and complex interclass relationships. The existing pure Markov chain approach for simulating multinomial classes has an apparent deficiency—underestimation of small classes, which largely impacts the usefulness of the approach. The Markov chain random field (MCRF) theory recently proposed supports theoretically sound multi-dimensional Markov chain models. This paper conducts a comparative study between a MCRF model and the previous Markov chain model for simulating multinomial classes to demonstrate that the MCRF model effectively solves the small-class underestimation problem. Simulated results show that the MCRF model fairly produces all classes, generates simulated patterns imitative of the original, and effectively reproduces input transiograms in realizations. Occurrence probability maps are estimated to visualize the spatial uncertainty associated with each class and the optimal prediction map. It is concluded that the MCRF model provides a practically efficient estimator for simulating multinomial classes from grid samples.     

安徽省新石器和夏商周时代遗址时空分布与人地关系的初步研究

安徽省新石器早期遗址有淮北的小山口和江南的沟汀,该时期淮河流域的聚落多于长江流域。新石器中期,考古发掘的野生动物骨骼和孢粉分析表明聚落周围是一个河湖密布、森林草原并存的生态环境。新石器晚期,古人已能分辨土壤的优劣并很可能会利用环境条件改善土壤的不良特性;蒙城尉迟寺遗址的剖面分析表明,环境只是古聚落分布和数量变化的基础因素,生产力的因素也在起作用,因为长时间的持续降温和夏初的大洪水侵袭。夏商周时期全省聚落总数增加,但淮北平原因为长时间的持续降温和夏初的大洪水侵袭,聚落大规模缩减。皖中和皖南的丘陵山地地貌使那里所受洪灾较轻,灾后这些地区因热量条件好使聚落迅速发展起来。新聚落大多建在利于农业生产、能躲避洪水的地方。