Incremental Semi-automatic Correction of Misclassified Spatial Objects

摘要

This paper proposes a decision tree based approach for semiautomatic correction of misclassified spatial objects in the Austrian digital cadastre map. Departing from representative areas, proven to be free of classification errors, an incremental decision tree is constructed. This tree is used later to identify and correct misclassified spatial objects. The approach is semiautomatic due to the interaction with the user in case of inaccurate assignments. During the learning process, whenever new (training) spatial data becomes available, the decision tree is then incrementally adapted without the need to generate a new tree from scratch. The approach has been evaluated on a large and representative area from the Austrian digital cadastre map showing a substantial benefit.