Research: Data capture

Ordnance Survey strives to remain at the forefront of collection and maintenance of spatial data. The Research Labs remote sensing team is researching means by which data can be collected from imagery and laser scanning data with superior efficiency, precision and accuracy. This is achieved by improving our current processes and products, or by investigating the processes we may use to capture future products. The research is strongly associated with work from the Data Modelling and Geousers teams. We are also collaborating with other parts of Ordnance Survey and external workers to ensure that this work is completed using state-of-the-art techniques.

To ensure that our mapping stays up to date, we need to identify change and update our database where and when change has occurred. We obtain intelligence about change from many sources, including imagery. Detecting change can be done by manually comparing images to our database, but we would like to make this a more automatic process.

Members of our change detection research team are investigating several methods, using various image classification and surface model comparison techniques. Ideally we would like to compare the results of the classification to our vector data (OS MasterMap), to identify changes to topographic features, especially buildings. Alternatively, we compare both images and DSMs from two different dates to detect any significant changes. Part of the research is to determine which changes are significant to us, and how we can identify these changes from the many other types of change which do not interest us (for example, a thick tree canopy in one image may be a mass of bare branches in the other - although this indicates a change, it is not one in which we are particularly interested).

Automatic change detection will ensure that the currency of our topographic products can be maintained more effectively in the future.

We are developing a 'capture once, use many' process to increase efficiency of capture and maintenance and the quality of our data products. This research takes theories of data integration to production-scale trials involving the capture of terrain height, topography and imagery data. 

By improving the efficiency of our data capture processes and by increasing the quality of the input data we are developing methods by which terrain height can be more accurate and more detailed. In turn, an high quailty terrain model ensures high quality orthorectification of imagery, which is a natural spin-off from our research.

The Digital Mapping Camera now used by Ordnance Survey has increased the information that can be derived from our imagery. In addition, we are investigating the contribution that lidar data could make to building the integrity of our data products.

We are investigating methods by which a 3D topography product can be captured both using manual and more automatic means. Photogrammetry and lidar are being investigated a sources for 3D data. We are focusing on buildings and invstigating the range of possible products that can be captured for these, from block model to highly detailed 3D data.

So that we are able to verify the quality of the data that we capture, we are developing accuracy measures for 3D data that are relevant to customer requirements.

Ordnance Survey have been researching the capture and processing of laser scanning (lidar) data since 2001. Research has focused on the processing of airborne laser data for the creation of terrain and building models.

More recently, this work has been extended to include investigations from mobile terrestrial platrforms and from waveform sensors. These new datasets are being used to enhance terrain and building models, and also to investigate the possibility of mapping vegetation and topographic attributes (such as slope roughness). The aim of the whole research programme is to identify the aspects of this data capture technique which may be of future benefit to Ordnance Survey.

Relevant papers 

SARGENT, I.; HARDING, J.and FREEMAN, M., 2007. Data quality in 3D: Gauging quality measures from users’ requirements. 5th International Symposium on Spatial Data Quality, Enschede.

AMIRI PARIAN, J. and SARGENT, I., 2007. Automatic height attribute assignment for building polygons: City modeling with level of detail zero 8th Conference on Optical 3-D Measurement Techniques, Zurich.

SANCHEZ HERNANDEZ, C., GLADSTONE, C. and HOLLAND, D., 2007. Classification of Urban Features from Intergraph's Z/I Imaging DMC high resolution images for integration into a change detection flowline within Ordnance Survey. Urban Remote Sensing Workshop, Paris.

HOLLAND, D. A. and TOMPKINSON, W., 2003. Improving the update of geospatial information databases from imagery using semi-automated user-guidance techniques. Geocomputation. Southampton.

TOMPKINSON, W., SEAMAN, E., HOLLAND, D. and GREENWOOD, J., 2003. An automated change detection system: Making change detection applicable to a production environment. MultiTemp, Ispra, Italy.

SMITH, S. L., HOLLAND, D., and LONGLEY, P., 2003. Interpreting Interpolation: The Pattern of Interpolation Errors in Digital Surface Models Derived from Laser Scanning Data. GISRUK. London.

SMITH, S. L., HOLLAND, D., and LONGLEY, P., 2003. The effect of changing grid size in the creation of laser scanner digital surface models. Geocomputation. Southampton.

SMITH, S. L., HOLLAND, D. and LONGLEY, P., 2003. Investigating the spatial structure of error in digital surface models derived from laser scanning data. 3D reconstruction from laser scanning. Dresden, Germany.