Leibniz MMS Days 2017 - Abstract

Schorcht, Martin

Spatial Segmentation Algorithm for Stratified Random Sample considering geometric circumstances

We developed a Segmentation Algorithm, which is based on Region Growing, for Stratified Random Sample. But instead of raster data our algorithm is using vector data! By the use of accurate vector data we can easily implement barriers, like rivers or railways, to control the growing we need to. Also by splitting polygons we achieve a kind of density gradient to slow the grow in unwanted regions, like bridges. As well we had to remove a lot of bottlenecks to reduce runtime and make it performant and practicable. The result are approximately optimal segmented regions, which are met certain conditions regarding the sample design and take consideration to geometric circumstances. In our case it is useful for reduction of costs on Germany-wide field survey for random sample of non-residential building, but is also applicable to any other wanted segmentation.