Bitte wählen Sie ihr Lieferland und ihre Kundengruppe
The three year project 'Programming techniques in remote sensing' has been funded by the Map Programme, Royal Norwegian Council for Scientific and Industrial Research with a total budget of NOK 2.55 mill. The primary aim has been to build up the competence and knowledge in handling remote sensing data, in particular for effective interpretation of data from high resolution satellites. The activities have been concentrated on contextual classification methods and development of a program system for raster image processing containing the implemented classification methods. The main results of the project, in addition to the general competence obtained by the members of the project group, are the development of A. A general supervised classification procedure incorporating 1) spatial dependcies between the true classes as well as the feature vectors of neighbouring pixels, 2) general loss functions and 3) options for rejecting pixels as doubters or outliers. B. A modular program system for raster image processing, running on a VAX 11/780 connected to a Ramtek Graphic Display System, particularly designed for the interpretation of data from high resolution satellites. C. A set of models describing the dependencies between the classes of adjacent pixels and the autocorrelation between the feature vectors. D. A set of methods for estimation of the parameters of the models.