Method for an automatic alignment of imagery and vector data applied to cadastral information in Poland

Abstract

Nowadays, an important problem in combining vector data and imagery is that they rarely align. This problem can become particularly acute in the case of cadastral systems. In this study, and as part of the partnership between the Universities of Jaén and Adam Mickiewicz (Poznań), we provide a methodological proposal to assess the conflation procedures between cadastral vector data and imagery, improving the alignment between both data sets. To do this, we use an automatic alignment algorithm which detects road intersections from both data sets as control points by using image texture characterisation. With this method, we first train the system on the imagery to learn the road texture distribution, then we can obtain its segmentation according to its texture, and finally the system locates road intersection points. The last step is to align vector data and imagery by using different techniques. This algorithm is based on an earlier one, detailed in [Ruiz, J.J., Rubio, T.J., and Ureña, M.A., 2011b. Automatic extraction of road intersections from images in conflation processes based on texture characterization. Survey review, 43 (321), 212–225.]. However, in the updated version we have solved the problem of not-well-defined intersection points, resulting in a substantial increase in the number of intersection points employed for the final adjustment to align both products and in a reduction of the computation time. On the other hand, the positional uncertainty assessment of parcel boundary lines both before and after applying our alignment procedure between them is provided. With regard to the experimental results, in the case of Polish cadastral data this procedure allows for significant improvement in the alignment between imagery and cadastral parcels boundaries.