Border Detection for Seamless Connection of Historical Cadastral Maps
Lecture Notes in Computer Science (2021)
This paper presents a set of methods for detection of important features in historical cadastral maps. The goal is to allow a seamless connection of the maps based on such features. The connection is very important so that the maps can be presented online and utilized easily. To the best of our knowledge, this is the first attempt to solve this task fully automatically. Compared to the manual annotation which is very time-consuming we can significantly reduce the costs and provide comparable or even better results. We concentrate on the detection of cadastre borders and important points lying on them. Neighboring map sheets are connected according to the common border. However, the shape of the border may differ in some subtleties. The differences are caused by the fact that the maps are hand-drawn. We thus aim at detecting a representative set of corresponding points on both sheets that are used for transformation of the maps so that they can be neatly connected. Moreover, the border lines are important for masking the outside of the cadastre area. The tasks are solved using a combination of fully convolutional networks and conservative computer vision techniques. The presented approaches are evaluated on a newly created dataset containing manually annotated ground-truths. The dataset is freely available for research purposes which is another contribution of this work.