Research project T4/DD/41 (Research action T4)
This research project aims to improve techniques of remote sensing image processing for land cover mapping and their integration into a geographic information system. The quality and the updating of these inventories are to be the basis of a sustainable and efficient land management (theme B.2.).
The project is motivated by the absence of an effective tool capable of processing the spatial information of remote sensing images and of using ancillary data structured into a geographic information system during the land cover interpretation. From then on, operational programmes of land cover cartography and inventory are essentially based on visual interpretation techniques. The updating of these data, soon necessary, will not be feasible in an effective manner unless the expertise acquired during the interpretation is used again. In addition, new sensors will allow in a near future to map land cover at scales lower than 1:25000, and therefore to detail current inventories.
From then on, it is proposed to develop a method and a software improving the recognition and the interpretation of land cover, in order to improve the use of spatial information and ancillary data during the interpretation of multi-sensors and multi-sources data, to preserve classification rules, and to stock the expertise resulting from the interpretation in order to use it again (updating or more detailed inventory on a larger scale).
The project started with a first two years phase. The first phase corresponded to the adaptation and to the validation of a classification software by decision tree to the land cover interpretation of remote sensing and ancillary data structured in a multi-sensors and multi-sources geographic information system. This method has the advantage of providing explicit classification rules and being adaptable according to the evolution of techniques. Results are interesting as they show that the adopted method is able to process a large number of different types of attributes at the same time, and that the resulting classification accuracy does not vary with attributes selection (i.e. a new attribute does not reduce the accuracy; if not necessary, it is not used).
At the end of the first phase, the main output is a software of assistance to the land cover recognition and interpretation from remote sensing data, under the form of a demonstrator, portable on PC, interactive, upgradeable, explicit (classification rules), cumulative (accumulation of the expertise) and adaptable (to new data).
Considering the first phase as successful, a second phase would consist in the improvement of the method developed by:
- increasing the abstraction level of image processing for classification,
- improving methods of land cover spatial generalisation,
- using this expertise in order to update a land cover inventory (recent high resolution data) and,
- refining a land cover inventory on a larger scale (recent very high resolution data).
This software has been developed by a trans-sectorial team associating cartographers and engineers in image processing. In the second phase, the network will be extended to a team specialised into artificial intelligence. The project has been guided by preferential relationship with two expected end-users implied in the inventory, the monitoring and the mapping of landcover: the European Topic Centre CORINE Land Cover and the National Geographic Institute of Belgium. These collaborations will be maintained in a second phase.
