Inproceedings,
Efficient area aggregation by combination of different techniques
Proceedings of the 10th ICA Workshop on Generalisation and Multiple Representation, August 2--3, 2007, Moscow, Russia, (2007)
Abstract
When reducing the scale of a topographic database, some areas of the data set become too small for representation and need to be aggregated with others, unintentionally but unavoidably leading to changes of some areas’ land cover classes. In this paper, we approach this problem by optimisation: Given a planar subdivision containing areas of different land cover classes, the problem is to aggregate areas into contiguous regions and to define the class for each region, such that each region satisfies a size threshold and the overall class change is minimal. A second objective is to create compact shapes. In an earlier paper we proved that the problem is NP-hard. We have presented a method by mixed-integer programming for its solution and introduced several heuristics. Our tests revealed that, even with the defined heuristics, our method does generally not allow to solve problem instances of more than 400 areas. In this paper we present a new efficient heuristic for the problem. Our approach is to locally introduce intermediate levels of details. Steps between these scales can be processed using our previously presented method. This approach allows processing large data sets – a complete map sheet of the German topographic map at scale 1:50.000 was processed to meet the requirements for the scale 1:250.000. We show that our method generalizes an existing iterative algorithm for the same problem and compare the results being obtained with different settings of our method. Compared with the existing iterative algorithm, our method resulted in 27,4% less change of land cover classes.
