Contextual classification

"Programming techniques in remote sensing" was a project funded by NTNF and carried out at NR in the period 1982-1984. The principal scope of the project was classification algorithms. A number of traditional as well as new techniques were implemented and tested. Some of these methods are contextual. Contextual classification is not only based on information carried by the pixel being classified, but on information from the neighbouring pixels as well. Experiments with the new techniques showed a significant improvement of classification accuracy. Methods developed in this project are now part of the methodological basis for the remote sensing activities at NR.

Remote sensing of forest

The possible use of satellite images in forestry was examined by NR in cooperation with Norwegian Forest Research Institute (NISK). The results showed that satellite imagery are unsuitable for the economic classification of forests by cutting class, but satellite images may still be applied in forestry. Mapping of clear cuttings and the growth of newly planted forest areas are candidates for satellite monitoring. A coarse taxonomy based on satellite images could also provide valuable resource information in developing countries, where the monitoring of forest and logging by traditional means often is inadequate.

Monitoring water quality

A project designed to determine whether the Landsat Thematic Mapper is suitable for monitoring water quality was carried out by NR in cooperation with Norwegian Institute for Water Research (NIVA) and Norwegian Hydrotechnical Laboratory (NHL). Water quality measurements were collected by boat while the satellite image was acquired. The results showed high correlation between image data and water quality parameters. The experiments have been continued by NIVA leading to operational monitoring.

Snow coverage in mountain areas

Knowledge of the amount of snow left in the upper watersheds is important information to the hydro-electric power companies. NR provided Statkraft (the largest Norwegian hydroelectric power company) with analyses of images from the AVHRR sensor in the NOAA weather satellite in order to determine the snow coverage throughout the melting season. The results are used as input data to snow melting simulation models. New techniques and methods are under development in this field, and NR has developed an operational semi-automatic system ran by Statkraft.

Classification of sea ice

Norway has strong scientific and economic interests in the Arctic. Radar satellites can be used to make observations in this region independent of the weather conditions. Classification of different categories of ice is one of the tasks radar images are useful for. The results may be used for scientific purposes in ice models and models of climatic change, and the results also have economic significance pertaining to shipping and to Arctic installations. Analysis of radar images obtained during SIZEX and MIZEX experiments was carried out by The Nansen Environmental and Remote Sensing Center (NERSC) and NR in order to decide whether automatic classification of ice categories was possible. The experiments showed that four different categories of ice could be discriminated.

Ice mapping in polar areas

Up-to-date ice maps showing the location and the movement of the ice would highly benefit shipping in the Arctic. Such maps are provided today, but these are low resolution maps updated and distributed only once a week. Weather independence and higher accuracy are possible by using radar images obtained by the radar satellite. NR has developed a prototype of a monitoring system designed to detect ice concentration, ice edge and ice motion from ERS SAR images. The system was also designed to present the results as thematic maps. The project was carried out under the authority of Norwegian Space Center (NRS).

Keeping maps uptodate is a well known problem in cartography. The Norwegian Mapping Authority is responsible for updating the national topographic maps (1:50,000 series) in Norway. The objective is to update all map sheets during a period of ten years. This has turned out to be impractical to accomplish. Today the updating cycle is close to 20 years. It may be possible to speed up the process by employing satellite images. The low spatial resolution of non-military satellites makes it impossible to gain the accuracy obtained from aerial photos. However, maps updated by coarse features could be published between the main revisions. Changes detected from the satellite image would be presented using dedicated graphical symbols in order to highlight the higher uncertainity attached to them. Maps revised by satellite images would present better information than old, non-revised maps. A prototype system based on semi-automatic map revision from satellite images was developed by NR and the Norwegian Mapping Authority. The system is able to detect changes in the road network and built-up areas by comparing a digital version of the old map to a recent satellite image. This kind of system would also be useful in countries where regular updates are absent.

Land-use statistics for urban areas

The Ministry of Environment has monitored five locations over the last 30 years in order to examine trends in the development of urban areas. Land use statistics is sampled on a 100 x 100 meter grid every 5th year, and the results are analysed statistically. A total of 200,000 samples are classified according to three different taxonomic schemes: area type (land use category), land parcel and physical object/structure. The work has so far been carried out by manual sampling from aerial photographs. The Norwegian Mapping Authority and NR have performed an experiment with automatic image analysis in order to reduce the manual work load. The experiment showed that automatic recognition of some of the classes is possible. A partial automation will bring about considerable savings in this work.

Monitoring soil erosion

Under the North Sea agreement, Norway had undertaken to cut the discharge of nutrient salts to the North Sea by 50% by 1995. One of the main sources of pollution is nutrient salts from fields tilled in the autumn. The government wanted to reduce the tillage in the autumn in regions vulnerable to erosion. In order to follow up on the initiative, the State Pollution Authority (SFT) wanted monitor soil tillage in areas drained to the North Sea. Observations from satellite seem to be one of the most cost-effective methods for this kind of monitoring. In cooperation with Norwegian Institute of Land Inventory (NIJOS) and Jordforsk (a soil research institute), NR carried out an experimental project in order to determine whether this is possible, and how satellite monitoring can be implemented as an operational activity. The experiments showed that fields can be classified as plowed, autumn grain, harrowed, stubble and pasture by Landsat TM images, provided that more than one image are used in the classification, and that the images are acquired at the optimum time during the autumn. An extension of the project includes an experiment where ERS-1 radar images were used. This experiment was justified from the fact that it may be difficult to obtain cloud-free images during the autumn in Norway, and it was carried out in cooperation with the Norwegian Defence Research Establishment (NDRE). The results showed that it is possible to distinguish between two classes, provided that the soil has a high moisture content: Ploughed and heavily harrowed fields constitute one class, all other fields the second class. Consequently, ERS and similar radar satellites can be used to monitor ploughing and coarse harrowing late in the autumn.

For more information, see this history document (in Norwegian)