Anvendelser (kun på engelsk)

Anvendelser (kun på engelsk)

A selection of application areas, where NR has improved or developed new methodology, is presented in the following. References are also provided to key projects where the research and development has taken place.

Monitoring environment and climate:

Natural resource mapping and monitoring:

Detection and mapping of man-made objects:

 

Monitoring climate change in the cryosphere: Observations of climate change indicators and climate model scenarios indicate that the global climate will change with increasing speed in the coming years with one of the most significant changes being a global warming. Europe is maybe the most sensitive region of the world and global warming most likely will change the living conditions in Europe significantly. NRs contributions to remote sensing methodology for cryospheric monitoring are mainly related to snow and monitoring system development. New algorithms have been developed for a new era of multi-sensor monitoring: multi-sensor time-series retrieval of snow cover area (Solberg, Malnes et al. 2005) and snow wetness (Solberg, Amlien et al. 2004). Furthermore, Keys algorithm has been adapted for retrieval of snow temperature (Amlien and Solberg 2003). The methods were demonstrated on Terra MODIS and ENVISAT ASAR data. From long-term observations of such variables, climate-change indicators can be derived, and climate models can be forced and validated against them. NR has co-ordinated the CEC FP5 project EuroClim (http://euroclim.nr.no) developing algorithms and a monitoring system doing this for snow, sea ice and land ice (Solberg 2004b). The results are applied by organisations like Norwegian Meteorological Institute, Max Plank Institute for Meteorology and European Environment Agency. The greatest current challenge to cryospheric monitoring by remote sensing is improvement of the accuracy of the geophysical variables. In order to be able to retrieve a very precise signal from a changing climate, also at the local and regional scale, further algorithmic improvements are needed urgently.

Detection of marine oil spill pollution: Oil spills are major sources of marine pollution. The amount of oil spill from rinsing tankers and "natural losses" in the Mediterranean alone is estimated to 600,000 tons yearly (three times the Amoco-Cadiz pollution). The problem is correspondingly large along the European Atlantic coast and in the Baltic. Marine oil spill can be detected by means of optical and radar sensors. Since SAR sensors are independent of sunlight and penetrate clouds, these are usually preferred for monitoring purposes. NR represents one of the leading research groups in the world with respect to development of automatic oil spill detection algorithms. The methodological approach was developed in the beginning of the 1990s and demonstrated on ERS-1 SAR data over the North Sea (Solberg and Solberg 1996). The results were brought into the CEC FP4 project ENVISYS, where a risk management system for oil spills was developed and demonstrated in the Mediterranean (Solberg and Theophilopoulos 1997). The algorithm was improved and tailored to ENVISAT ASAR and Radarsat data in the CEC FP5 project Oceanides. A comparative analysis of manual, semi-manual and automatic approaches represented in the project was also carried out, which showed that the automatic approach could compete with the manual and was better than the semi-automatic (Solberg 2004b). New challenges for improving the oil spill detection rate include improvement of the algorithms assignment of oil spill confidence levels, inclusion and handling of ancillary data and data from other sensors, and inclusion of operator feedback for retraining the algorithm.

Monitoring the state of forests and natural vegetation: Forests and mountain vegetation comprise a dominating part of the natural vegetation in Norway. The traditional monitoring paradigm, based on field surveys, is expensive and error prone. Recent advances in air- and space-borne monitoring technology has gradually opened to more efficient and reliable approaches for monitoring the state of the natural vegetation. NR contributed to environmental monitoring of forests in the CEC PF5 project FOREMMS developing a system prototype for long-term pan-European monitoring (see Fjrtoft and Solberg 2002). Recently, NR has cooperated with Skogforsk in health monitoring of forests based on SPOT imagery, see Solberg, Lange et. al. (2005) and Solberg, Nsset et. al. (2005). NR also has significant activities related to mountain vegetation within the CEC FP6 project geoland where grazing land issues and vegetation change is addressed.

Mapping snow for improved water management: The seasonal snow cover is practically limited to the northern hemisphere. In the mountainous areas and in the whole north of Europe, snowfall is a substantial part of the overall precipitation. Monitoring the seasonal snow is important for several purposes. In northern regions, the snow may represent more than half the annual runoff, putting specific demands on the models and other tools employed in managing this water resource. Risk of flooding strengthens this demand, both in areas with stable winter coverage, and in areas only occasionally covered with snow. Optical remote sensing sensors are able to map snow cover area (SCA) quite accurately, but are limited by clouds. Synthetic Aperture Radar (SAR) sensors penetrate the clouds, but current satellite-borne sensors are only able to map wet snow accurately. NR has contributed to the development of new and innovative algorithms at the research frontier for more than a decade. This work has taken part in several projects; in particular in the Research Council project SnowMan and the CEC projects SnowTools in FP4, and EnviSnow and EuroClim in FP5. New algorithms have been developed for single-sensor optical SCA retrieval (Solberg 2005a), multi-sensor time-series SCA retrieval (Solberg, Malnes et al. 2005), snow surface wetness (SSW) retrieval (Solberg, Amlien et al. 2004) and retrieval of snow temperature at surface (STS) (Amlien and Solberg 2003). The methods were demonstrated using NOAA AVHRR, Terra MODIS, Envisat ASAR/MERIS/AATSR and Radarsat data. The current major challenges in remote sensing of snow is to further develop the new optical SCA algorithm for retrieval in forested areas as well and to harmonise the retrieved information from optical and SAR data in multi-sensor time-series algorithms for SCA and SSW.

Forest resource mapping: The forestry industry is a major economic sector in Norway. Mapping and inventorying of the forest resources has been an important activity in Norway for centuries. The classical approach to forest mapping relies heavily on fieldwork by forestry specialists. Manual interpretation of stereoscopic aerial imagery has for a long time been an integral part of this fieldwork. New optical satellites like SPOT, with resolutions better than 2.5 m, are making satellite-based forest resource mapping feasible. NR has a long tradition in remote sensing of forest resources, see for example Strand (1989) and Aas et al. (1996).The focus of these projects has been to either determine the tree type or to retrieve forest cutting class. There are important problems related to image interpretation of optical imagery, especially in regions with steep topography where tree height and local altitude variations are hard to distinguish. The use of a detailed DEM is necessary in such regions. The need for sensor fusion has arisen with the advent of new sensor technology.

Mapping mountainous grazing land: The use of remote pastures for feeding domestic animals is a longstanding tradition in Norway. The land and right owners have an economic interest in estimating the grazing value in these regions, and such interests may also include hunting. National and local authorities also need to know the grazing value in order to take actions when pastures are over-utilized. NR has worked with these issues within the CEC FP6 project Geoland, which addresses land cover and vegetation in general. Links have been established with Norwegian users and with various European research entities working with remote sensing in nature management and land cover mapping. Research contributions from NR include classification of multi-temporal data utilizing phenological models (Aurdal et al. 2005a, Huseby et al. 2005a). The challenge of deriving the grazing maps more directly from earth observation data should be approached by first establishing a suitable set of aggregated vegetation classes that can be linked to grazing quality. Retrieval methods may involve interpretation and knowledge injection from ancillary data.

Updating maps of new and changed infrastructure: Changes in societal infrastructure, such as buildings, roads, train tracks, airports, harbours etc., happen at an ever-increasing rate. In order to keep up to date with these changes, frequent map revisions are necessary. Such revisions are however very costly. A possible alternative to the largely manual map revision procedures common today is based on semi- or fully automatic feature extraction from aerial and satellite imagery. Modern satellites, such as Quickbird, provide optical images with spatial resolutions as high as 0.5 meters, largely sufficient for many infrastructure-monitoring problems. This resolution can be increased further by using airborne sensors. The key technologies for feature extraction from such imagery are image processing and pattern recognition. Applications of image processing and pattern recognition for feature extraction from aerial and satellite imagery are currently the subjects of intensive research. Research within these fields has a long tradition at NR, see for instance (Robb and Solberg 1988; Solberg et al. 1990; 1991). Digitization of existing map data has also been an important research topic at NR, see (Eikvil et al. 1995a; 1995b). NR is currently collaborating with a private company for the development of a new generation of map revision techniques. The main challenges that must be overcome in order to improve the quality of current systems for infrastructure monitoring are related to how models of infrastructure can be used to interpret actual observations in digital imagery. This is a complex problem requiring that the model be used to explain image data, at the same time as the image data are used to refine the model.

Vehicle detection for traffic statistics: Vehicle locations and movements are important indicators of the capabilities of the infrastructure to serve transport needs. Such information is important for national transport planning and accessibility analyses, as well as analysis of environmental effects. Fixed installations, like inductive loops and stationary cameras, are currently used to acquire traffic flow information. These installations provide information at one point in space over a long time period. A valuable complement to this is a complete spatial view of the traffic, providing road traffic information acquired at one point in time over a large area. Such views are particularly important for road networks in larger cities, where this type of snapshot information cannot be acquired by other means. Satellites, like Ikonos and Quickbird, can now provide images with a resolution of 0.5-1 meter, opening up for applications aiming at vehicle detection and traffic monitoring based on satellite images. NRs contribution to this area is a clear understanding of relevant techniques, acquired through many years of research in land-based remote sensing applications and pattern recognition for other applications. In addition to this comes a good knowledge of the area of traffic information from work with statistical models for car traffic and automatic traffic monitoring based on video analysis (Eikvil et al. 2001).

Detection of potential cultural heritage sites in agricultural fields: The presence in the ground of deteriorating remains of, for instance, a prehistoric building is likely to modify local ground chemistry. Observed in aerial or satellite imagery of an agricultural field, such a site might therefore be visible as a region with different properties when compared to its surroundings. There is currently great interest from Norwegian cultural heritage conservation authorities in mapping possible sites for such prehistoric buildings, roads etc. This field is also internationally the focus of intensive research. The primary focus is research of the use of aerial and satellite images for the purpose of documenting known sites and for inspecting (manually) images in order to find new sites. Since 2001 NR has been involved in work related to interpretation of aerial and satellite imagery in order to identify possible cultural heritage sites (Aurdal 2003; Aurdal et al. 2005b). The aim of this project is to produce a system that can be used by local authorities for cultural heritage site detection. The main challenge in the development of such a system is the classification of spots in agricultural fields. The spots are to be classified as interesting or uninteresting locations, the problem obviously being that the variability can be very great in either class. In this work NR draws heavily on experience gained in classification of amorphous objects for oil spill detection.

Innsatsområde