3D-Stadtmodellierung auf heterogenen Ausgangsmodellen mittels evolutionaerer Algorithmen

This thesis introduces a new approach for flexible and efficient 3D city modeling. The approach establishes a comprehensive and consistent target model out of a set of existing heterogeneous basic models. To this aim, first a unified data pool is built from the set of basic models. Then the target model is derived automatically from the pool by transforming the basic models and detecting their relations. The basic models may vary widely with respect to resolution, quality, coverage and data format. Moreover, the basic models may even be partially contradictory to one another. Thus, the derivation of a target model is a complex optimization problem. Evolutionary algorithms (EA), known from the discipline of computer science, are an up-to-date method to resolve these problems.The basic idea of EA is to transfer mechanisms inspired by Darwin’s theory of evolution. All data sets are treated as individuals, which may reproduce, mutate, recombine and die. By using an adequate “survival of the fittest“ evaluation system the individuals are converging towards a target model. To apply EA in our context, a data model has to be specified that satisfies the requirements of EA, the basic models and the 3D city modeling. Based on the current state of the art in both EA and 3D city modeling, the EA-model is analyzed, adapted and extended. The data sets converge towards the target by operators, which realize EA basic calculations. A set of operators is designed. Their single and combined effects are analyzed in the given context. The results are demonstrated with practical examples. The approach can be generalized towards overlapping, heterogenous data sets.