Automatic positional accuracy assessment of geospatial databases using line-based methods

Abstract

In this paper, we present a methodology for automating the positional assessment of vector geospatial databases (GDBs). We establish a framework which works with a non-punctual perspective of the accuracy assessment problem, i.e. considering polygons (which represent buildings) as closed linear shapes and using techniques based on buffer generation on the perimeter-lines, which allow us to analyse the displacement between the location of the polygons stored in a GDB (tested source) and a location determined by another GDB (reference source) having higher accuracy. In order to determine a set of homologous objects of both GDBs, a genetic algorithm has been used. The results obtained demonstrate the viability of this evaluation and its potential compared to traditional methods.