Adaptive Zoning for Transport Mode Choice Modeling

Adaptive zoning is a recently introduced method for improving computer modeling of spatial interactions and movements in the transport network. Unlike traditional zoning, where geographic locations are defined by one single universal plan of discrete land parcels or ‘zones’ for the study area, adaptive zoning establishes a compendium of different zone plans, each of which is applicable to one journey origin or destination only. These adaptive zone plans are structured to represent strong spatial interactions in proportionately more detail than weaker ones. In recent articles, it has been shown that adaptive zoning improves, by a large margin, the scalability of models of spatial interaction and road traffic assignment. This article confronts the method of adaptive zoning with an application of the scale and complexity for which it was intended, namely an application of mode choice modeling that at the same time requires a large study area and a fine‐grained zone system. Our hypothesis is that adaptive zoning can significantly improve the accuracy of mode choice modeling because of its enhanced sensitivity to the geographic patterns and scales of spatial interaction. We test the hypothesis by investigating the performance of three alternative models: (1) a spatially highly detailed model that is permissible to the maximum extent by available data, but requires a high computational load that is generally out of reach for rapid turnaround of policy studies; (2) a mode choice model for the same area, but reducing the computational load by 90% by using a traditional zone system consisting of fewer zones; and (3) a mode choice model that also reduces the computational load by 90%, but based on adaptive zoning instead. The tests are carried out on the basis of a case study that uses the dataset from the London Area Transport Survey. Using the first model as a benchmark, it is found that for a given computational load, the model based on adaptive zoning contains about twice the amount of information of the traditional model, and model parameters on adaptive zoning principles are more accurate by a factor of six to eight. The findings suggest that adaptive zoning has a significant potential in enhancing the accuracy of mode choice modeling at the city or city‐region scale.     

Alternatives to current GPS-RTK services and some implications for CORS infrastructure and operations

The justification for the establishment of CORS networks was initially in support of geodesy and other geoscientific applications at the global and regional level. However, increasingly GPS CORS network operators have sought ways of making their network infrastructure the basis of a profitable business. This has arisen with the introduction of real-time centimeter-level accuracy services, carrier phase-based modes of operation generally referred to as GPS-RTK (real-time kinematic). One approach is to try to recruit a core group of users who are prepared to pay for the GPS-RTK services. But this is only feasible if the number of users, and the fees that are charged, are sufficient to generate a reasonable return-on-investment (ROI). This ROI (or at the very least “cost-recovery”) is important for many network operators in order that they may provide for the maintenance and upgrade of the CORS infrastructure. On the other hand, there are those who advocate that there is no need to recoup CORS investment, that the installed GPS receivers should be seen as public infrastructure in a similar manner to roads, bridges, etc. This paper discusses some new business and operational models for GPS-RTK services. These include models for the establishment and operation of CORS infrastructure, service provision, business cases, and options for value-added services beyond the standard GPS-RTK service. One concept is based on a “client–server” model. Currently GPS-RTK service providers have no control over the quality of the results computed by users. This makes it difficult for them to justify charging for their services. What if instead of broadcasting RTK corrections and placing the onus of obtaining a final solution on the user and his equipment, the user’s coordinates are determined by the service provider? Putting the computational effort on the server side will justify more easily the charging of users for a value-added product: an accurate and quality assured coordinate in the local reference frame. This paper describes the client–server concept as well as possible business models that may underpin such a service model. These models include some derived from mobile telephony and service/hospitality businesses. Furthermore, with the projected proliferation of independent, competitive GPS-RTK services, the concept of a GPS data or service “broker” is worth exploring.     

面阵数码航摄像控测量布设的思考

对在像控测量布设中遇到的2个问题进行了阐述,得出航线网精度估算式仍然适用,但数码航摄仪的像控点数量应多于光学航摄仪的结论。另外,在满足一定的条件下可进行跨航线布点,像控点的选点目标需要顾及成图精度要求,不能照搬现有规范要求。     

多结构卷积神经网络特征级联的高分影像土地利用分类

高分辨率遥感影像包含丰富的土地利用类型信息，针对单一卷积神经网络提取图像特征信息不足的问题，提出了一种多结构卷积神经网络（convolutional neural network，CNN）特征级联的分类方法。首先，选择CaffeNet（convolutional architecture for fast feature embedding）、VGG-S（visual geometry group-slow）、VGG-F（visual geometry group-fast）为实验初始模型，对网络全连接层进行参数微调，采用随机梯度下降法（stochasticgradient descent，SGD）更新网络的权重；然后以微调后的网络分别作为特征提取器对图像提取特征，级联上述3种网络的第二个全连接层输出特征作为图像表达；最后，以多类最优边界分配机（multi-class optimal margindistribution machine，mcODM）获得最终分类结果。实验采用UC Merced land-use数据集进行分类效果检验，结果表明，多结构卷积神经网络级联的方法能够达到97.55%的总体分类精度，相较于CaffeNet、VGG-S和VGG-F等，分类精度分别提升了5.71%、2.72%和5.1%。因此多结构卷积神经网络特征级联的方法能够有效提取目标特征信息，提升土地利用分类精度。     

Open source map matching with Markov decision processes: A new method and a detailed benchmark with existing approaches

Map matching is a widely used technology for mapping tracks to road networks. Typically, tracks are recorded using publicly available Global Navigation Satellite Systems, and road networks are derived from the publicly available OpenStreetMap project. The challenge lies in resolving the discrepancies between the spatial location of the tracks and the underlying road network of the map. Map matching is a combination of defined models, algorithms, and metrics for resolving these differences that result from measurement and map errors. The goal is to find routes within the road network that best represent the given tracks. These matches allow further analysis since they are freed from the noise of the original track, they accurately overlap with the road network, and they are corrected for impossible detours and gaps that were present in the original track. Given the ongoing need for map matching in mobility research, in this work, we present a novel map matching method based on Markov decision processes with Reinforcement Learning algorithms. We introduce the new Candidate Adoption feature, which allows our model to dynamically resolve outliers and noise clusters. We also incorporate an improved Trajectory Simplification preprocessing algorithm for further improving our performance. In addition, we introduce a new map matching metric that evaluates direction changes in the routes, which effectively reduces detours and round trips in the results. We provide our map matching implementation as Open Source Software (OSS) and compare our new approach with multiple existing OSS solutions on several public data sets. Our novel method is more robust to noise and outliers than existing methods and it outperforms them in terms of accuracy and computational speed.     

Simplifying the interpretation of continuous time models for spatio-temporal networks

Autoregressive and moving average models for temporally dynamic networks treat time as a series of discrete steps which assumes even intervals between data measurements and can introduce bias if this assumption is not met. Using real and simulated data from the London Underground network, this paper illustrates the use of continuous time multilevel models to capture temporal trajectories of edge properties without the need for simultaneous measurements, along with two methods for producing interpretable summaries of model results. These including extracting ‘features’ of temporal patterns (e.g. maxima, time of maxima) which have utility in understanding the network properties of each connection and summarising whole-network properties as a continuous function of time which allows estimation of network properties at any time without temporal aggregation of non-simultaneous measurements. Results for temporal pattern features in the response variable were captured with reasonable accuracy. Variation in the temporal pattern features for the exposure variable was underestimated by the models. The models showed some lack of precision. Both model summaries provided clear ‘real-world’ interpretations and could be applied to data from a range of spatio-temporal network structures (e.g. rivers, social networks). These models should be tested more extensively in a range of scenarios, with potential improvements such as random effects in the exposure variable dimension.

     

Crowdsourcing the character of a place: Character‐level convolutional networks for multilingual geographic text classification

This article presents a new character‐level convolutional neural network model that can classify multilingual text written using any character set that can be encoded with UTF‐8, a standard and widely used 8‐bit character encoding. For geographic classification of text, we demonstrate that this approach is competitive with state‐of‐the‐art word‐based text classification methods. The model was tested on four crowdsourced data sets made up of Wikipedia articles, online travel blogs, Geonames toponyms, and Twitter posts. Unlike word‐based methods, which require data cleaning and pre‐processing, the proposed model works for any language without modification and with classification accuracy comparable to existing methods. Using a synthetic data set with introduced character‐level errors, we show it is more robust to noise than word‐level classification algorithms. The results indicate that UTF‐8 character‐level convolutional neural networks are a promising technique for georeferencing noisy text, such as found in colloquial social media posts and texts scanned with optical character recognition. However, word‐based methods currently require less computation time to train, so currently are preferable for classifying well‐formatted and cleaned texts in single languages.     

城市道路复杂度的浮动车自适应采样算法

基于浮动车的城市交通信息采集技术是智能交通系统获取实时交通信息的重要手段之一。针对浮动车的城市交通信息等间距采样的不足,本文设计了一种基于城市道路复杂度的自适应采样算法:①根据道路属性定义道路结点对城市道路网络复杂度的影响因子;②利用四叉树对城市道路网络复杂度进行描述;③根据浮动车的瞬时速度和道路复杂度自适应计算浮动车的采样周期。通过仿真和试验表明,新算法能够在不同复杂程度的道路情况下自适应提供有效、可靠的采样周期。     

An adaptive compromise programming method for multi-objective path optimization

Network routing problems generally involve multiple objectives which may conflict one another. An effective way to solve such problems is to generate a set of Pareto-optimal solutions that is small enough to be handled by a decision maker and large enough to give an overview of all possible trade-offs among the conflicting objectives. To accomplish this, the present paper proposes an adaptive method based on compromise programming to assist decision makers in identifying Pareto-optimal paths, particularly for non-convex problems. This method can provide an unbiased approximation of the Pareto-optimal alternatives by adaptively changing the origin and direction of search in the objective space via the dynamic updating of the largest unexplored region till an appropriately structured Pareto front is captured. To demonstrate the efficacy of the proposed methodology, a case study is carried out for the transportation of dangerous goods in the road network of Hong Kong with the support of geographic information system. The experimental results confirm the effectiveness of the approach.     

Simulating the effects of social networks on a population’s hurricane evacuation participation

Scientists have noted that recent shifts in the earth’s climate have resulted in more extreme weather events, like stronger hurricanes. Such powerful storms disrupt societal function and result in a tremendous number of casualties, as demonstrated by recent hurricane experience in the US Planning for and facilitating evacuations of populations forecast to be impacted by hurricanes is perhaps the most effective strategy for reducing risk. A potentially important yet relatively unexplored facet of people’s evacuation decision-making involves the interpersonal communication processes that affect whether at-risk residents decide to evacuate. While previous research has suggested that word-of-mouth effects are limited, data supporting these assertions were collected prior to the widespread adoption of digital social media technologies. This paper argues that the influence of social network effects on evacuation decisions should be revisited given the potential of new social media for impacting and augmenting information dispersion through real-time interpersonal communication. Using geographic data within an agent-based model of hurricane evacuation in Bay County, Florida, we examine how various types of social networks influence participation in evacuation. It is found that strategies for encouraging evacuation should consider the social networks influencing individuals during extreme events, as it can be used to increase the number of evacuating residents.     

Follow the road: historical GIS for evaluating the development of routes in the Negev region during the twentieth century

ABSTRACT

At the beginning of the twentieth century, a British mapping team led by Captain S. F. Newcombe surveyed and mapped the Negev region, Sinai, and western Jordan. The map was mainly produced for military use. Consequently, it included a network of branched routes, water supplies and facilities, and topographic contours. This study used this map to examine the development of routes in the Negev region between the beginning of and until the end of the twentieth century. First, the individual sheets comprising the study area were pieced together and the accuracy of the map was evaluated. The accuracy found on the Newcombe map was 0.76 mm on the map scale, equivalent to 100.3 m. Route development during the twentieth century was then evaluated by comparing the routes digitized from the Newcombe map to digitized routes on a late twentieth-century map. The results do not reveal tremendous changes in path, shape, or number of routes. Instead, they merely indicate the natural development in their quality. This Historical GIS-based approach provided a useful technique for analyzing and comparing the line segments extracted from historical and modern maps. The implemented approach may also serve other geographical or historical studies aiming to examine the development of branched networks throughout history.     

网络空间同位模式的加色混合可视化挖掘方法

同位模式挖掘是空间数据挖掘的热点问题之一,应用领域广泛。已有的同位模式挖掘方法一般采用统计或数据挖掘的方式,要求对复杂的数学公式、算法及相关参数等有深刻的理解,主要针对同质的欧式空间中地理现象。而城市空间中人为地理现象大多发生在网络空间,鉴于此,本文提出了一种网络空间同位模式可视化挖掘方法。该方法利用视觉语言表达网络空间现象之间的影响和交互作用。首先,利用网络空间核密度估计表达网络空间现象的分布情况和影响范围,为网络空间现象的同位模式挖掘提供支持,并建立单个地理现象分布情况与颜色之间的映射;然后基于色光加色混合原理获得两个地理现象相互影响的认知,借以挖掘空间同位模式。本文提出的方法属于形象思维,具有直观,形象和易感受等特点。     

Bayesian networks and agent-based modeling approach for urban land-use and population density change: a BNAS model

Land-use change models grounded in complexity theory such as agent-based models (ABMs) are increasingly being used to examine evolving urban systems. The objective of this study is to develop a spatial model that simulates land-use change under the influence of human land-use choice behavior. This is achieved by integrating the key physical and social drivers of land-use change using Bayesian networks (BNs) coupled with agent-based modeling. The BNAS model, integrated Bayesian network–based agent system, presented in this study uses geographic information systems, ABMs, BNs, and influence diagram principles to model population change on an irregular spatial structure. The model is parameterized with historical data and then used to simulate 20 years of future population and land-use change for the City of Surrey, British Columbia, Canada. The simulation results identify feasible new urban areas for development around the main transportation corridors. The obtained new development areas and the projected population trajectories with the“what-if” scenario capabilities can provide insights into urban planners for better and more informed land-use policy or decision-making processes.     

曲线多尺度表达的Head-Tail信息量分割法

本文提出一种基于Head-Tail信息量分割的地理要素多尺度表达模型。首先通过傅里叶变换将地理线要素转换为傅里叶描述子,并通过香农信息熵理论计算其频域信息量。然后,按Head-Tail数据分布模式确定地理要素信息量的分界点,并设计函数对各个分界点所对应的傅里叶截断频率进行估计。最后,参考传统方根模型,建立以地理要素频率信息量为基础的信息方根模型,计算与各个地图层次相对应的关键尺度,实现地理要素的层次化多尺度表达。采用等高线及海岸线的数据试验表明,本文所提出的模型能够有效根据设定的比例尺对地理要素进行化简,对不同目标比例尺的简化结果体现出了良好的区分度与层次性。同时,在保证化简结果与原地理要素面积重叠比一致的情况下,本文模型所得到的结果优于传统的简化算法。     

利用LSTM网络预测月球物理天平动参数

利用中国探月甚长基线干涉测量（very long baseline interferometry, VLBI）观测数据改进月球物理天平动参数的预测精度,对于着陆器和巡视器的精密定位具有重要意义。利用VLBI单点定位模型解算得到“嫦娥三号”（Chang’E-3,CE-3）着陆器的坐标和物理天平动,分别采用循环神经网络（recursive neural network, RNN）和长短期记忆（long-short term memory, LSTM）网络进行物理天平动的预测。选取月球着陆器的坐标和VLBI观测量作为输入量,将3个欧拉角Ω, i, μ作为输出量,将11 323个样本用于训练,2 315个样本用于测试,2 315个样本用于验证,1 000个样本用作与预测结果进行对比。结果显示, 验证集的数据经过1 000次训练和9次迭代训练后的梯度约为6.2×10-5(″)/s,证明了LSTM网络与RNN的可靠性。LSTM网络和RNN的3个欧拉角的预测精度分别达到了97.8%、99.7%、97.2%和95.2%、98.5%、95.8%,LSTM网络的预测精度更高。与DE421星历对欧拉角的预测结果进行比较,结果证明了LSTM网络预测精度更高。     

A Hybrid Link‐Node Approach for Finding Shortest Paths in Road Networks with Turn Restrictions

Turn restrictions, such as ‘no left turn’ or ‘no U‐turn’, are commonly encountered in real road networks. These turn restrictions must be explicitly considered in the shortest path problem and ignoring them may lead to infeasible paths. In the present study, a hybrid link‐node Dijkstra's (HLND) algorithm is proposed to exactly solve the shortest path problem in road networks with turn restrictions. A new hybrid link–node labelling approach is devised by using a link–based labelling strategy at restricted nodes with turn restrictions, and a node‐based labelling strategy at unrestricted nodes without turn restrictions. Computational results for several real road networks show that the proposed HLND algorithm obtains the same optimal results as the link‐based Dijkstra's algorithm, while having a similar computational performance to the classical node‐based Dijkstra's algorithm.     

多目标微观邻域粒子群算法及其在土壤空间优化抽样中的应用

提出一种基于多目标微观邻域粒子群的土壤空间优化抽样方法。方法面向土壤空间调查的多目标特征,构建了基于最小克里金方差（MKV）和极大熵准则（ME）的粒子群多目标适应度函数,设计了最小样本量限制、样点可达性、采样成本限制和最小空间关联性四类粒子微观邻域操作策略,能高效协调土壤空间抽样精度、代表性、成本、样本量与样点布局等多目标冲突。实验结果表明,相比单目标粒子群算法和模拟退火算法,该方法的目标冲突协同能力强、收敛效率高,所设计抽样方案最优,为土壤质量精确调查与高效监测提供了技术支持。     

基于增强DeepLabV3网络的高分辨率遥感影像分类

由于高分影像具有地物细节丰富、类别差异大等特点,现有的卷积神经网络影像分类方法普遍存在分类精度低、地物边界不准确等问题。鉴于此,本文提出一种基于增强DeepLabV3网络的影像分类模型。首先构建R-MCN网络结构,利用大小不同的卷积核并结合残差网络的思想进一步提取浅层网络的多尺度、多层次的特征信息;然后采用可学习的上采样方式,并将R-MCN提取的特征与高阶的语义信息相融合;最后通过提出的Mloss损失函数,获得遥感影像的地物分类结果。试验结果表明,相对于传统的卷积神经网络,本文方法能细化地物的边缘信息,改善分类效果,获得更高的影像分类精度。     

深度学习支持下的地图图片典型地理目标检测

针对地图图片中典型地理目标识别问题,本文首先介绍了两种基于深度学习的目标检测方法(YOLO网络和采用focal loss替换交叉熵损失函数的RetinaNet网络),然后将地图图片分别输入两种神经网络模型中进行训练和测试,最后对目标检测结果进行对比分析。结果表明,RetinaNet网络模型对地图图片进行目标检测的准确率有明显提高,且运行速度依然可达秒级。该地理目标检测方法的高准确度与高效性可在地图审查时节约大量人力、时间成本,为地图内容智能理解及互联网地图监管提供了新的技术参考。     

自动制图综合人工神经元网络方法的研究

现代神经生理学研究表明 ,人脑的大量神经元构成有所分工而又紧密联系的神经元网络 ,它的结构和功能可以采用物理可实现的系统人工神经元网络来模拟。制图综合是人脑神经元网络获取、处理、输出地理信息的复杂视觉思维过程 ,可以用人工神经元网络来模拟。文中探讨了制图综合的人工神经元网络的设计 ,并对用于实现自动制图综合的结果进行分析 ,指出其应用前景。