Sitemap Contact Kalender Nieuw Home

Projectendatabank FEDRA





The GIS problem detector (GISproblem)

Onderzoeksproject S0/02/032 (Onderzoeksactie S0)

Personen :

Beschrijving :

Context and objectives

Tele Atlas is the market leader in the production of digital road maps in a wide variety of applications all over Europe and the USA. The company is faced with the problem of upgrading its current databases to a higher accuracy and ensuring the quality of the information. Current techniques can’t support this in a cost-effective way due to the necessary manpower. Automated detection of change and anomalies in the existing databases using very high resolution (VHR) satellite images can form an essential tool to support quality control and maintenance of spatial information.
Quality of geospatial data is still an evolving concept. At first glance, it seems a simple concept but devising measures which in a meaningful way quantize different discrepancies that can occur in the data is not straightforward. Two questions will be dealt with within the project:

• To what extent can remote sensing and computer vision be used to detect deviation of a GIS database vs. “real world”. The target is to have a high reliability level on the reported data problems.
• In a second phase, these parameters and algorithms will be used to design an automatic quality assessment system to produce a quality statement of GIS databases.


Expertise in GIS as well as image interpretation are combined to cover the problem of quality control from different angles, i.e. a top-down approach starting from the vector data, establishing how quality can be defined and measured, and a bottom-up approach starting from the raster source data, defining the information that can be reliably extracted from images. These two approaches combined allow us to develop a realistic and feasible system which can be integrated within a production environment. The work is split up in 3 parts:

• Information extraction from VHR image data: Interesting features such as corners and ridges are extracted. The performance of the extraction is characterized using error propagation.
• Quality assessment of geospatial data: Based on the available literature, some quality measures are proposed and evaluated for the given imagery and road vector database.
• Automated quality assessment using VHR data: The extracted image information will be used to perform the developed quality assessment procedure in an automated way, detecting problem areas in the road vector database


In the first part, two ridge detectors are introduced. The performance of the road extraction is characterized in terms of the detection rate and mean segment length. The results show that there is a good correspondence between the predicted and the measured detector performance. This enables us to set up the detector parameter set in such way an optimal extraction of the road features can be expected. As a result, a prototype has been created to extract road features out of various image sources without the need of an image processing expert. Experiments have been conducted on a set of images of typical roads in an IKONOS satellite image to verify the validity of the derived expressions.
In the second part of the research, a thorough analysis of the literature has been conducted concerning quality assessment of linear vector databases. We set up a framework for applying the selected procedure, i.e. Buffer Overlay Statistics (BOS), on both vector- and raster-layers. Using this methodology we did an assessment of the quality of the existing TeleAtlas-database. We can conclude that the average displacement of the TA-data in the considered test zone is 1,8m.

To conclude, we set up a generic framework to automatically assess the quality of a road network database based on very high resolution images. Two approaches were tested starting from reliable and unreliable georegistration of the imagery. The first approach used the developed Image-based Buffer Overlay concept, the second approach error tolerant graph matching.

Documentatie :

Over deze website

Uw privacy

© 2017 POD Wetenschapsbeleid