Data acquisition: Meeting demand from space

Prof. Ian Dowman
Department of Civil, Environmental and Geomatic Engineering
University College London
London (UK)
idowman@cege.ucl.ac.uk

Images from space-based platforms are now established as a useful and economic source of data for extracting information for mapping and geospatial data bases. The images which can be used for this purpose range from sensors such as MODIS with 250m, 500m and 1000m pixel size and wide global coverage with a short time frame, Landsat with 60m, 30m and 15m pixel size giving regular medium resolution coverage for monitoring, to high resolution images from sensors with 0.5m pixel size, with the ability to access any part of the globe within 24 hours but only giving irregular global coverage.

Another important group of sensors collects stereoscopic images for generation of digital elevation models (DEMs). The data also includes data from microwave sensors using synthetic aperture radar technology, now producing resolution of 1m, and being used interferometrically to give elevation data and to monitor tectonic movement and subsidence at the millimeter level. These imaging sensors are supported by Global Navigation Satellite Systems (GNSS) which are vital for providing positional information, both for mapping directly and for giving the position of other platforms. The combination of imaging system and positioning system, together with the use of inertial navigation systems, has been the driving force behind many important developments in recent years; particularly LiDAR, mobile mapping systems and interferometric synthetic aperture radar systems (IfSAR). GNSS is also essential for positioning satellite imaging sensors and has enabled images to be acquired with high georeferenced accuracy.

Enhanced image resolution
The most important trend over the past 46 years has been the improvement in image resolution. Table 1 illustrates the trend to higher resolution. It also shows anomalies relating to particular applications. The SPOT satellites for example give data continuity over 24 years at 10m pixel size; MOMS was a research sensor, maybe ahead of its time, SPOT HRS is designed for generating digital elevation models.


Along with improved resolution having improved georeferencing, SPOT 1 data was geolocated to kilometer level whilst this is now down to metres for the most recent sensors. These developments have enabled organisations producing geospatial information to use the data commercially, or for national mapping and also for the data to be an important source of information for applications such as disaster management.

Most of the satellites included in table 1 are not classified as small satellites. Sandau (2008) defines a small satellite as being less than 1000kg and lists those currently flying. The most significant issue with small satellites is their development by countries other than those with large space agencies: Nigeria, Algeria, Korea, China (Taipei), Korea being particular examples and many of these have been built by Surrey Satellite Technology Limited (SSTL). SSTL has also initiated the disaster monitoring constellation (DMC) which pools resources of the members for rapid response to disasters.


Stereoscopic data acquisition
Another important development has been the improved acquisition of stereoscopic data. Some sensors such as ASTER, SPOT HRS and ALOS PRISM allow the collection of two or three images within seconds of each other, using fore and aft sensors, whilst the agile sensors such as Worldview and GeoEye can collect stereo data in a single overpass by changing the pointing direction of the single sensor.

Sensors using interferometric SAR have also been used for this purpose, the prime example being SRTM (Shuttle Radar Topography Mission). SRTM has produced a near global DEM with 1 arc second spacing and elevation precision of better than 16m. A more recent development has been the ASTER GDEM.

As well as the improved resolution, these high resolution sensors all have multispectral channels, providing colour images which are co-registered with the panchromatic channel. Data fusion software allows these channels to be combined in whatever way suits the application. A popular product is an orthoimage, which has relief distortion corrected, but requires a digital elevation model (DEM) in order to remove the effect of the terrain, but a 3D model with buildings to remove the distortion caused by buildings.

3D data creation
Three dimensional data can be generated using stereo imagery and manual mapping. An important development has been the use of rational polynomial coefficients to create 3D models on a workstation. This removes the need for users to enter a specific sensor model, or for control for orientation, although control is still necessary for checking the orientation, but it also requires the data provider to give the coefficients with the image. RPCs were viewed with some suspicion initially, but are now widely used without problems.

Lower resolution multispectral sensors have been developed for specific applications, for example MODIS and Landsat, but their application is generally for scientific research. The applications of this high resolution satellite data are very varied. Perhaps the widest use of the data is for applications such as Google Earth and Bing, where uncorrected data is used to populate global data sets. Global mosaics of MODIS and Landsat data are of lower resolution, but better geolocated, and can be used for scientific purposes. DEMs are used to produce orthoimages which generally give much better location accuracy.


Use of radar
The more radical development has been the use of radar from space. Sensors such as ERS 1 and 2 and Radarsat 1 were designed to generate synthetic aperture radar (SAR) images for various environmental and intelligence applications, but it was found that the ERS data, which was provided with high accuracy positional information, could be used to generate interferograms which could be used to generate high accuracy elevation data, and by using a third image, measure movement due to tectonic activity or subsidence. Radar sensors have continued to improve with Envisat continuing from ERS, Radarsat 2 and recently TerraSAR- X and Cosmo-skyMed. Improvement in interfeometry has also continued. SRTM is a notable example, but TerraSAR-X is also used for this and will be continued with TanDEM-X.

Constellation launch
The most recent development in sensor technology has been the launching of constellations. The Disaster Monitoring Constellation (DMC) has been operating for some time using the sensors produced by SSTL in UK. Rapid Eye is a prime example of this, being a constellation of five sun-synchronous Earth observation providing large area, multi-spectral images with frequent revisits with 6.5m ground sampling distance (GSD) and orbital position knowledge of less than 10 meters with 95% confidence. The Cosmo-skyMed constellation consists of 4 mediumsize satellites, each one equipped with a microwave highresolution SAR operating in X-band, with resolution of up to 1m. The Group on Earth Observation (GEO) and the Committee on Earth Observation Satellites (CEOS) is investigating a new constellation to fill in gaps in the Earth observation programme required to satisfy the data requirements of the Global Earth Observing Systems of Systems (GEOSS).

Future trends
In terms of resolution for the future, it is well known that better resolution of optical sensors is possible, but restricted by issues of national security, so new development will come from policy decisions which will allow higher resolution and wider access to data. There appears to be a sound commercial market for satellite data and there is still room for improvement in the level of acceptance of such data for 3D mapping. Improved software for feature extraction and image analysis is still required, for 3D mapping automatic extraction of buildings and roads is a critical requirement and still a major research activity. The use of small satellites still has to prove itself as a cost efficient tool for smaller nations, but consortia such as African Resources Management (ARM) Satellite Constellation may be able to make good use of this type of technology.

Space agencies such as ESA, NASA, NOAA have long term programmes for collecting data for environmental studies, with a wide range of specialist instruments. Organisations such as the Group on Earth Observation (GEO) will have an influence on the further development of data acquisition from satellite platforms by studying gaps in provision of data for important global requirements such as climate change, management of water and forests and understanding changes in the natural ecology and biodiversity.

So in summary we can say that space-based data acquisition is in a strong position, providing products for which there is a significant demand, for commercial use, national mapping and scientific research. The available technology still has scope for improvement in resolution and data processing, but development of this will depend on policy decisions relating to resolution and data access.