Qi Zhou

Integration of linear and areal hierarchies for continuous multi-scale representation of road networks

Spatial data can be represented at different scales, and this leads to the issue of multi-scale spatial representation. Multi-scale spatial representation has been widely applied to online mapping products (e.g., Google Maps and Yahoo Maps). However, in most current products, multi-scale representation can only be achieved through a series of maps at fixed scales, resulting in a discontinuity (i.e., with jumps) in the transformation between scales and a mismatch between the available scales and users desired scales. Therefore, it is very desirable to achieve smoothly continuous multi-scale spatial representations. This article describes an integrated approach to build a hierarchical structure of a road network for continuous multi-scale representation purposes, especially continuous selective omission of roads in a network. In this hierarchical structure, the linear and areal hierarchies are constructed, respectively, using two existing approaches for the linear and areal patterns in a road network. Continuous multi-scale representation of a road network can be achieved by searching in these hierarchies. This approach is validated by applying it to two study areas, and the results are evaluated by both quantitative analysis with two measures (i.e., similarity and average connectivity) and visual inspection. Experimental results show that this integrated approach performs better than existing approaches, especially in terms of preservation of connectivity and patterns of a road network. With this approach, efficient and continuous multi-scale selective omission of road networks becomes feasible.

A comparative study of various strategies to concatenate road segments into strokes for map generalization

The study of road networks has been a topic of interest for some time. A road network in a database is often represented by intersections and segments. However, in many cases (e.g., traffic flow analysis and map generalization), one needs to consider individual roads as a whole, instead of individual segments. Thus, it is sometimes very desirable to concatenate road segments into long lines – ‘strokes’ as they are called in the literature. For stroke building, a number of strategies are available and the effectiveness of using these strategies needs to be evaluated. This article presents a comparative analysis of 17 such strategies, including 3 of the geometric approach, 1 of the thematic approach, and 13 of the hybrid approach for road network generalization purposes. Three sets of real-life data with different patterns are used to test these strategies. Corresponding road maps at smaller scales are used as benchmarks and a new measure called the accuracy rate is proposed to indicate the correctness of the concatenated strokes. The results show that if only the geometric approach is considered, the every-best-fit strategy performs best; if thematic attributes are also added, road class can be more effective than road name. Also significance tests (the chi-square test and the Marascuilo procedure) are carried out to give all pairwise comparisons of these strategies. The results indicate that 45 of the 136 pairs of strategies have statistically significant differences; the purely geometry-based every-best-fit performs significantly better than the purely geometry-based self-fit; and the inclusion of thematic attributes, especially road class, sometimes improves the accuracy rate but the improvement is not significant.

Use of Artificial Neural Networks for Selective Omission in Updating Road Networks

Abstract An important problem faced by national mapping agencies is frequent map updates. An ideal solution is only updating the large-scale map with other smaller scale maps undergoing automatic updates. This process may involve a series of operators, among which selective omission has received much attention. This study focuses on selective omission in a road network, and the use of an artificial neural network (i.e. a back propagation neural network, BPNN). The use of another type of artificial neural network (i.e. a self-organizing map, SOM) is investigated as a comparison. The use of both neural networks for selective omission is tested on a real-life road network. The use of a BPNN for practical application road updating is also tested. The results of selective omission are evaluated by overall accuracy. It is found that (1) the use of a BPNN can adaptively determine which and how many roads are to be retained at a specific scale, with an overall accuracy above 80%; (2) it may be hard to determine which and how many roads should be retained at a specific scale using an SOM. Therefore, the BPNN is more effective for selective omission in road updating.

Empirical determination of geometric parameters for selective omission in a road network

ABSTRACT Selective omission in a road network is a necessary operation for road network generalization. Most existing selective omission approaches involve one or two geometric parameters at a specific scale to determine which roads should be retained or eliminated. This study proposes an approach for determining the empirical threshold for such a parameter. The idea of the proposed approach is to first subdivide a large road network, and then to use appropriate threshold(s) obtained from one or several subdivisions to infer an appropriate threshold for the large one. A series of experiments was carried out to validate the proposed approach. Specifically, the road network data for New Zealand and Hong Kong at different scales (ranging from 1:50,000 to 1:250,000) were used as the experimental data, and subdivided according to different modes (i.e. administrative boundary data, a regular grid of different sizes, different update years, and different road network patterns). Not only geometric parameters, but also structural and hybrid parameters of existing selective omission approaches were involved in the testing. The experimental results show that although the most appropriate thresholds obtained from different subdivisions are not always the same, in most cases, the appropriate threshold ranges often overlap, especially for geometric parameters, and they also overlap with those obtained from the large road network data. This finding is consistent with the use of different subdivision modes, which verifies the effectiveness of the proposed approach. Several issues involving the use of the proposed approach are also addressed.

How many samples are needed? An investigation of binary logistic regression for selective omission in a road network

ABSTRACT Selective omission in a road network (or road selection) means to retain more important roads, and it is a necessary operator to transform a road network at a large scale to that at a smaller scale. This study discusses the use of the supervised learning approach to road selection, and investigates how many samples are needed for a good performance of road selection. More precisely, the binary logistic regression is employed and three road network data with different sizes and different target scales are involved for testing. The different percentages and numbers of strokes are randomly chosen for training a logistic regression model, which is further applied into the untrained strokes for validation. The performances of using the different sample sizes are mainly evaluated by an error rate estimate. Significance tests are also employed to investigate whether the use of different sample sizes shows statistically significant differences. The experimental results show that in most cases, the error rate estimate is around 0.1–0.2; more importantly, only a small number (e.g., 50–100) of training samples is needed, which indicates the usability of binary logistic regression for road selection.

Experimental Analysis of Various Types of Road Intersections for Interchange Detection

Road interchanges are a major pattern type in road networks. Recognition of road interchanges benefits automated road network generalization, car navigation and traffic flow analysis. This study first investigated several existing approaches to automatically detecting interchanges in a road network, and determined that the recognition of characteristic road intersections is essential for the effective detection of interchanges. Several experiments were carried out to investigate nine types of road intersections for interchange detection and to analyze characteristic ones. Furthermore, an approach to the detection of both intersections and segments of interchanges was proposed and validated. The road networks across different scales were tested and results show that: (1) the T-shaped and Cross-shaped junctions are very common in road networks, but they are not the most characteristic ones for interchange detection; (2) the y-shaped, Y-shaped types, X-shaped, Fork-shaped and Multi-leg may be the characteristic types for interchange detection; (3) the proposed approach to detecting interchanges is effective, and most of the intersections and segments of interchanges can be detected. In addition, taking multiple characteristic types into consideration for interchange detection is suggested.

Evaluation of Properties to determine the Importance of individual Roads for Map Generalization

Many researchers have paid much attention to the importance of individual roads for various applications such as traffic flow analysis and road network generalization. This paper gives a comparative analysis of different properties to determine the importance of individual roads for road network generalization purpose. These properties include one geometric property (length); three topological properties (degree, closeness and betweenness) and one thematic property (road class). Two representative selective omission approaches (stroke-based selection and mesh densitybased elimination) are implemented to generalize road network. For each approach, different properties are respectively used to determine the importance of individual roads. The road network of Hong Kong Island is used as study area for testing and two measures (similarity and connectivity) are employed to evaluate the selections. Results show that, when the stroke-based selection approach is implemented, using length performs best in determining the importance of individual roads, and using betweenness or closeness performs well in preserving the connectivity of the retained network; when the mesh density-based elimination approach is implemented, all these properties have quite similar selections.

A Comparative Study of Various Supervised Learning Approaches to Selective Omission in a Road Network

Selective omission is necessary for road network generalisation. This study investigates the use of supervised learning approaches for selective omission in a road network. To be specific, at first, the properties to measure the importance of a road in the network are viewed as input attributes, and the decision of such a road is retained or not at a specific scale is viewed as an output class; then, a number of samples with known input and output are used to train a classifier; finally, this classifier can be used to determine whether other roads to be retained or not. In this study, a total of nine supervised learning approaches, i.e., ID3, C4·5, CRT, Random Tree, support vector machine (SVM), naive Bayes (NB), K-nearest neighbour (KNN), multilayer perception (MP) and binary logistic regression (BLR), are applied to three road networks for selective omission. The performances of these approaches are evaluated by both quantitative assessment and visual inception. Results show that: (1) in most cases, these approaches are effective and their classification accuracy is between 70% and 90%; (2) most of these approaches have similar performances, and they do not have any statistically significant difference; (3) but sometimes, ID3 and BLR performs significantly better than NB and SVM; NB and KNN perform significantly worse than MP, SVM and BLR.

Exploring the relationship between density and completeness of urban building data in OpenStreetMap for quality estimation

ABSTRACT OpenStreetMap (OSM) is a free spatial data source based on crowd sourced data. Although the OSM data have a range of applications, such as generating 3D models, and routing and navigation, quality issues are still significant concerns when using the data. Several studies have undertaken quality assessments by comparing OSM data with reference data. However, reference data are not always available due to high costs or licensing restrictions, and very few studies have quantitatively estimated the quality of OSM data under conditions where the corresponding reference data are not available. This study proposed the use of a building density (or building coverage ratio) indicator as a proxy, and designed a series of experiments involving different study areas to quantitatively explore the relationship between building density and building completeness for OSM data in urban areas. The residuals (estimated building completeness and reference building completeness) were also analyzed. Two main results were found from the experiments. (1) There was an approximate linear relationship between building density and building completeness in the OSM data. More precisely, the building completeness of OSM data was approximately 3.4–4 times the building density of OSM data. (2) Approximately 70–80% of the absolute residuals were smaller than 10%, and 80–90% of them were smaller than 20%. This shows that, in most cases, estimated building completeness was close to the corresponding reference building completeness. Therefore, we concluded that the building density indicator is a potential proxy for the quantitative completeness estimation of OSM building data in urban areas. The limitations of using this indicator were also addressed.

Rethinking the Buffering Approach for Assessing OpenStreetMap Positional Accuracy

OpenStreetMap (OSM) is a free source of spatial data based on crowd-sourcing. Although OSM data are widely used in applications such as the generation of 3D models, routing and navigation, the quality issue is still one of the significant concerns when using these data. Extensive studies have focused on assessing the quality, especially the positional accuracy , of OSM data. One method for assessing accuracy is the buffering approach where a buffer is created around a validated road network using a predefined buffer radius. The percentage of OSM road lengths that lie within this buffer is then calculated. While existing studies have used the buffering approach, the method itself has not been evaluated either theoretically and experimentally. It is found that the percentage of OSM road length calculated based on the buffering approach may be imprecise if the validated road network and the OSM road network are not matched one-to-one. Therefore, this study suggests that it is necessary to first match the OSM road network with the validated road network before using the buffering approach.