Automatic delineation of built-up area at urban block level from topographic maps
Introduction
The analysis of urban form and urban morphology has a long history (Bianchin & Bravin, 2008: 300; Hofmeister, 2004: 5) and is of high importance for stakeholders of urban governance such as authorities of local and interurban spatial planning (Foltête and Piombini, 2007, Krumdieck et al., 2010), climate researchers (Heldens et al., 2010, Peeters and Etzion, 2012), or decision makers in economics (Webster, 2010). Ex-post analyses of built-up areas1 based on time series are most relevant for planning and evaluation processes in terms of achieving settlement policy goals (e.g. to reduce urban sprawl) (Herold et al., 2005, Jaeger et al., 2010). Further, the analysis results can be used to feed ex-ante models of urban dynamics with necessary auxiliary input data (Herold et al., 2005, Jantz et al., 2010). Both forms of urban analysis require data about the morphology of the built-up area ranging from (digital) topographic data to remote sensing data. The latter include aerial/satellite imagery (e.g. Taubenböck et al., 2010, Thiel et al., 2008, Wurm et al., 2010) and airborne laser scanning data (Yoshida & Omae, 2005). While recent research focuses on information retrieval from very high resolution remote sensing systems, data of earlier generations (e.g. Landsat TM/ETM +) lack the spatial resolution to allow small-scale analysis and are only suited for regional-scale delineation of built-up area (Taubenböck, Esch, & Racette, 2011).
To derive extensive historic information about urban morphology for ex-post analyses potential surveying methods vary. While most case studies are of local character and, therefore, are based on visual interpretation to demarcate built-up areas from open space (Herold et al., 2005, Jaeger et al., 2010), methodologies have been suggested to automatically extract this information (cf. in Stengele, 1995). Scanned topographic maps have the advantage of implicitly holding semantically structured information about the urban fabric that is spatially accurate within given restrictions due to generalization. This information, in order to be analyzed, needs to be converted to explicit spatial objects by means of pattern recognition and image processing (Graeff & Carosio, 2002). According to Walter and Luo (2011), this process can be referred to as map interpretation in general or inverse cartographic engineering in particular. Behind the background of a global trend to digitize knowledge that can also be observed with topographic maps (e.g. USGS, 2013), methods for an automated map interpretation form the necessary foundation to analyze the resulting geographic knowledge database. Previous research has shown the suitability of topographic maps for (ex-post) analyses of urban morphology (e.g. Meinel, Hecht, & Herold, 2009), despite their inherent level of uncertainty and generalization (Leyk et al., 2005, Tucci and Giordano, 2011).
Recently, high-resolution analysis of urban morphology focused on the development and application of building-based methods, e.g. to describe stock changes or to derive indicators of urban morphology relevant for planning (Carneiro et al., 2010, Meinel et al., 2009). However, the next superior spatial entity of the urban fabric – the urban block representing a group of developed parcels with buildings – remains an important reference for spatial planning and official areal statistics. The spatial segmentation of built-up space into urban blocks is a common concept of urban morphology. The urban block represents an important morphological unit that not only helps to globally differentiate forms of urban morphology based on quantitative characterization, but also allows for a comparative monitoring of urban metamorphism at city or urban district scale (Wurm et al., 2010, Yoshida and Omae, 2005).
In this paper, we propose an approach using methods of image processing and geometrical constraints to automatically extract urban information from historic topographic maps placing our emphasis on the primary objects of urban fabric: urban block and street network, both sharing a mutual relationship (Edwardes & Mackaness, 2000). While databases already exist for contemporary data (e.g. DLM), our approach is important to constitute historical data. We demonstrate the applicability of our methodology using binary German topographic maps at a scale of 1:25.000. Section 2 will give a short summary of related work in urban object extraction in general and map interpretation in context of urban analysis in particular. Section 3 covers the data input and the chosen study sites for testing the methodology that is presented in Section 4. The results for our study sites are given in Section 5 and discussed in an international context in Section 6. Section 7 will give a summary and present an outlook on implications to future work.
Section snippets
Object extraction from raster maps
In reference to urban analysis, the delineation of urban objects used to be mainly a task of visual interpretation of maps and aerial images. With emerging new technologies, the focus moved to an automated analysis of images showing urban phenomena (Bianchin & Bravin, 2008). Accompanying the paradigm shift from analog to digital cartography and the consequent building up of digital (historic) spatial databases, the automated image analysis of maps had its first peak some 20 years ago (e.g.
Topographic data
Derived from the scope of our research to allow for comprehensive monitoring of built-up area in time series, the approach for delineating urban blocks from historic topographic maps proposed here is based on the binary black layer of the DTK25(-V) — an abbreviation for the scan of a hard copy German topographic map at scale 1:25.000. It has a scan resolution of 508 dpi. Being at large comparable to other European maps, the black layer contains the elementary objects of urban morphology:
Methodology
Most of the widely accepted methods developed for delineating objects with linear demarcation from images or digitized maps solely use a raster-based approach to separate the object of interest from the ‘background’, including morphological filtering and thresholding for color images. Many of them appear to create good results depending on the quality of the input or the object of interest. The hereby presented concept, however, appends such methods by an object-based approach (raster-vector
Evaluation
The methodology has been trained based on three tiles of the size 1250 ∗ 1250 m for the map sheets and all map object models have been tuned to achieve perfect delineation results for those tiles. We have added a table of parameters and the assigned values used in each module as Supplementary online material. Due to the amount of parameters and depending on the map layout complexity, the adaption can be time consuming and requires some knowledge of image analysis, however. To evaluate the results
Discussions of results
The overall delineation result is very promising. However, for multi-fragmented or incomplete street block boundaries the approach shows only medium performance. This can be observed especially in industry-dominated map regions, areas with complex, superimposing transportation infrastructure, and inner city areas with an especially high competition of different map contents. Further, one particular problem that can be observed in peripheral settlement areas is the use of dashed rather than
Conclusion and outlook
We have presented an approach to automatically interpret topographic maps with regard to street blocks and built-up areas (urban blocks) that significantly reduces manual effort to prepare scanned historic maps for integration into a spatio-temporal database and urban analysis. Using a modular process that applies methods of image analysis and spatial analysis, we are able to overcome the challenges that are inherent in topographic maps concerning the automated delineation of built-up area.
Acknowledgments
This research is funded by the German Research Foundation (ME 1592/3-1, ME 1592/3-2). The authors greatly appreciate this funding. The data used to proof our concept are kindly provided by the Federal Office for Cartography and Geodesy (BKG).
References (55)
- et al.
Automatic identification of urban settlement boundaries for multiple representation databases
Computers, Environment and Urban Systems
(2008) - et al.
Urban layout, landscape features and pedestrian usage
Landscape and Urban Planning
(2007) - et al.
The role of spatial metrics in the analysis and modeling of urban land use change
Computers, Environment and Urban Systems
(2005) - et al.
Urban permeation of landscapes and sprawl per capita: New measures of urban sprawl
Ecological Indicators
(2010) - et al.
Designing and implementing a regional urban modeling system using the SLEUTH cellular urban model
Computers, Environment and Urban Systems
(2010) - et al.
Urban form and long-term fuel supply decline: A method to investigate the peak oil risks to essential activities
Transportation Research Part A: Policy and Practice
(2010) - et al.
Automated recognition of urban objects for morphological urban analysis
Computers, Environment and Urban Systems
(2012) - et al.
Positional accuracy, positional uncertainty, and feature change detection in historical maps: Results of an experiment
Computers, Environment and Urban Systems
(2011) - et al.
Automatic interpretation of digital maps
ISPRS Journal of Photogrammetry and Remote Sensing
(2011) Pricing accessibility: Urban morphology, design and missing markets
Progress in Planning
(2010)
An approach for analysis of urban morphology: Methods to derive morphological properties of city blocks by using an urban landscape model and their interpretations
Computers, Environment and Urban Systems
Remote sensing and urban analysis
A system for automatic extraction of road network from maps
Automatic street graph construction in sketch maps
Recognition of hatched cartographic patterns.
Digital urban morphometrics: Automatic extraction and assessment of morphological properties of buildings
Transactions in GIS
Conflation of geospatial data
A method for automatically extracting road layers from raster maps
A general approach for extracting road vector data from raster maps
International Journal on Document Analysis and Recognition
Automatic and accurate extraction of road intersections from raster maps
GeoInformatica
Alnwick, Northumberland: A study in town-plan analysis
Transactions and Papers
Knowledge-based segmentation for automatic map interpretation
Intelligent generalisation of urban road networks
Delineation of urban footprints from TerraSAR-X data by analyzing speckle characteristics and intensity information
IEEE Transactions on Geoscience and Remote Sensing
Automatic interpretation of scanned topographic maps: A raster-based approach
Image-based quality assessment of road databases
International Journal of Geographical Information Science
Cited by (18)
Computer vision applications for urban planning: A systematic review of opportunities and constraints
2024, Sustainable Cities and SocietyGeographic features recognition for heritage landscape mapping – Case study: The Banda Islands, Maluku, Indonesia
2023, Digital Applications in Archaeology and Cultural HeritageDeveloping urban built-up area extraction method based on land surface emissivity differences
2020, Infrared Physics and TechnologyCitation Excerpt :Landsat, National Polar-Orbiting Partnership Visible Infrared Imaging Radiometer Suite, Synthetic Aperture Radar, World view, and other satellite remote sensing data sources are used to extract UBA, and there are many effective methods, such as spectral index related to impervious surface area (ISA), combined spectral index, spectral unmixing model, land cover classification, etc.[3–9]. The second category is the use of statistical survey data, such as field survey and land planning data, geographical national conditions census data, topographic maps, and other UBA boundary data [10–12]. The third category is based on the application of artificial intelligence technology and big data, such as urban taxi global positioning system tracking data, combined index technology, and electronic map points of interest [13–16].
The MASi repository service — Comprehensive, metadata-driven and multi-community research data management
2019, Future Generation Computer SystemsCitation Excerpt :Through digital map collections curated by libraries and national mapping agencies (e.g. [6,7]) amounts up to several million digitalized old maps are made available. To efficiently conduct large-scale and long-term analyses on historical maps we developed georeferencing methods [8] and information extractions procedures (e.g. for buildings [4,5,9] and settlement areas [10,11]) that can be automated to a high degree. This (semi-)automatic processing generates and necessitates a variety of metadata.
Refining flood estimation in urbanized catchments using landscape metrics
2018, Landscape and Urban PlanningCitation Excerpt :The use of landscape metrics in hydrology has however been limited, despite showing promise in predicting urban land-use change impacts through representation of form and function (Lin, Hong, Wu, Wu, & Verburg, 2007; Van de Voorde et al., 2016). Comparatively, urban ecological research, which has long been using ecological typologies to study ecosystem dynamics (Brady et al., 1979), has evolved into many detailed landscape metrics of landscape structure in dedicated spatial statistical software (Kupfer, 2012) with diverse applications (e.g. Alberti, 2005; Jiao, 2015; Muhs, Herold, Meinel, Burhardt, & Kretschmer, 2016). Within ecological landscape metrics, distance is often considered as Euclidean and thus is not calculated according to a hydrological network.