Generalisation of 1:50k roads centrelines from 1:10k road polygons in an automated workflow

Abstract

This paper reports about on going developments in the map production process of the Dutch Kadaster (who also hold the mapping agency). In March 2013, the Dutch Kadaster will produce the 1:50k map series fully automatically from 1:10k data for the first time. From that moment on, this new product will replace the existing 1:50k map. It is the first time that a complete topographic map is generalised with no human interaction and that the resulting map is good enough to replace the existing map. The implementation is a 100% automated generalisation workflow that produces a countrywide map series at scale 1:50k from 1:10k data within 50 hours. Therefore a 1:50k map delivery is foreseen with every 1:10k data update, i.e. five times a year. The research that achieved the fully automated generalisation workflow was explained in several publications (Stoter et al, 2009; 2011; 2012). The automated workflow for the generalisation of 1:50k maps from TOP10NL data is currently being extended to produce the 1:100k map series. This workshop paper focuses on the generalisation problem for which one third of the sub models were developed: generalisation of the road network. The Dutch data models at 1:10k and 1:50k have two properties that characterise this process. At first 1:10k data contains road polygons for every road wider than 2 meters that represent the real widths of the roads, while 1:50k maps contain road centerlines that cover complete roads and that have an attribute indicating the width of the road. The second property that characterises the generalisation of roads is that data at all scale form a planar partition, i.e. no gaps or overlap are allowed. Therefore every polygon that is removed should be replaced by something else.