Optimization Methods for Area Aggregation in Land Cover Maps

The aggregation of areas is an important subproblem of the map generalization task. Especially, it is relevant for the generalization of topographic maps which contain areas of different land cover, such as settlement, water, or different kinds of vegetation. An existing approach is to apply algorithms that iteratively merge adjacent areas, taking only local measures into consideration. In contrast, global optimization methods are proposed in this paper to derive maps of higher quality. Given a planar subdivision in which each area is assigned to a land cover class, we consider the problem of aggregating areas such that defined thresholds are satisfied. The aggregation is directed at two objectives: Classes of areas shall change as little as possible and compact shapes are preferred. In this paper, the problem is formalized and two different approaches are compared, namely mixed-integer programming and simulated annealing.