Elaboration of methodology of generalisation of DTM with the use of artifi cial intelligence methods

Aim of project
Correct generalisation of DTM is particularly important for supplying Geographical Information Systems (GIS) with spatial data. In order to perform reliable spatial analyses it is particularly important to maintain real locations of characteristic points of key terrain features. Therefore, the generalisation process should be considered as the DLM model generalisation, and not as the DCM (digital cartographic model) generalisation. Therefore, automation of large scale DTM modelling and generalization, considered in this way, requires that modern algorithms of automatic generalisation, as for example, machine learning techniques, are applied.

Nowadays, the algorithmic approach may be considered as the dominating tendency in the fi eld of (model and cartographic) generalisation. It consists of utilization of strict, parameterised procedures of utilization of elementary generalisation operators: simplifi cation, aggregation, filtration etc. Results of utilisation of artifi cial intelligence and cognitive modelling in the process of generalisation of spatial data, are also very promising. On the contrary to expert systems, well-known since the eighties of the 20th century, which utilize IF-THEN deterministic rules, the essence of the discussed approach is connected with the use of machine learning (ML) processes.

The following algorithms are classified into the ML group of algorithms: fuzzy inference systems (FIS), rough sets, neural networks or decision trees.

The discussed methods are cognitive modelling tools which allow to perform data reduction, but, first of all, they allow for explanation of complexity of the systems which are modelled.

Expected results
The new methodology and technology of DTM models generalisation have been proposed, which utilize cognitive modelling and diversifi ed ML algorithms (neural networks, systems of fuzzy and neuro-fuzzy inference, as well as rough sets);

Generalisation results of DTM models obtained with the use of machine learning algorithms have been compared with results of conventional terrain surface generalisation;

The role of terrain surface model generalisation in the process of Geographical Information Systems (GIS) supply with data has also been discussed.