1 de julio de 2000 Vol. 1 No.1


INTEGRATED TECHNIQUE FOR AUTOMATED DIGITIZATION OF RASTER MAPS
Serguei Levachkine and Evgueni Polchkov

(continue...)

4 Raster Map Post-Processing

The main goal of the post-processing of raster maps (after cartographic image recognition) is the correction of automated vectorization errors.

For automated correction of raster map digitization we suggest two approaches: using the topological characteristics of objects in vector layers and the spatial correlation (connectivity) of the corresponding vector layers.

The first approach is based on the fact that many cartographic objects in the system have topological characteristics which can be used for the correction of vectorizing errors. Let us give some obvious examples.

Isolines. The topological characteristics of isoline systems (for example, contour lines) are: a) isolines are continuous, b) they cannot intersect each other, c) each isoline is either closed or starts and finishes at a domain boundary, and d) polygons which cover the whole domain without intersections can be assembled from the arcs of the correct isoline system together with the domain boundaries.

However on a raster map these characteristics, as a rule, may be lost for several reasons: a) the lines are broken where a parameter value for a given isoline is written, b) some sections of isolines are not well drawn in high density regions, and c) the raster images of some isolines merge due to defects of printing and scanning the paper maps. The "tick marks" (small segments of fixed length, which are perpendicular to the isolines and drawn in the decreasing direction of the cartographic parameter) need special consideration. These elements of the map's graphical design, if not recognized as the parts of the isoline system, hinder the correct assembly of the polygons and either must be eliminated or (better) detached in a separate vector layer. They can be restored on the vector map and used for the automated attribution of polygons assembled from the contour lines.

Hydrologic network. The relevant topological characteristic of the hydrological network (rivers, channels, brooks, etc.) is that its vector image, within the limits of each watershed, has the structure of a simple connected directed tree graph. However on raster maps, images of bodies of water can be disrupted by images of other types: areal (reservoir, cities, etc.) and point (populated places, bridges, etc.) that break the structure of the hydrologic network.

Region maps. The topological characteristic of a region map (for example, the map of political divisions of a country) is that each arc, being a segment of the boundary of a disjoint region, is either closed or starts and finishes at another arc. Thus a correct system of non-intersecting polygons covering the whole surface of the map can be assembled from the boundaries of the regions, adding, if necessary, the domain boundaries.

However on the vectored region map, the boundaries of some regions may be discontinuous, impeding the production of the correct map topology.

The second approach is newer and offers more potential. It consists in using connectivity, or spatial correlations among the various vector objects to correct the results of automated digitization and the attributes of raster map cartographic images. Let us explain this by the following examples.

Hydrosphere elements. The vectored elements of the hydrosphere include objects of all three topological types: polygonal (e.g. seas, lakes, reservoirs, ponds, bogs), linear (e.g. rivers, channels, brooks) and point (e.g. springs, out-of-scale reservoirs). It is clear that nearly all hydrosphere elements have certain spatial relations among one other. For example, springs or lakes can be the sources of rivers (main hydrosphere elements). Linear river sections can connect polygonal hydrosphere elements (extended sections of riverbeds, lakes or reservoirs). Rivers discharge into seas or lakes. These spatial relations are not only necessary for a topologically correct vector image of the hydrosphere, but can be also used to correct automated digitization results of the river network, where the digitized polygonal and point hydrosphere elements are available.

Relief. The relief of a region on the maps is represented by objects of two topological types; point (trigonometric points, hydrometric monitoring posts, reference points with coordinates obtained by the Global Positioning System (GPS)), and linear (contour lines, relief features–slopes, ravines, landslides, etc.). All these objects have correlations among each other due to their location and in particular to the altitudes given as attribute information.

We note particularly the relation between altitude and the hydrologic network; altitude decreases monotonically along a river in the direction of flow, thus each contour line intersects the river either not at all or exactly once. Due to this relationship, it is convenient to produce the vector map of the hydrological network before digitizing the contour lines and to subsequently use the digitized hydrological network for correction of the results of contour digitization.

Road network. The road network on the map represents roads, highways and railways of all sizes. Its elements form a graph, the majority of whose nodes represent populated places and point objects of the transportion infrastructure (e.g. bridges, tunnels, railway crossings, docks, airports) and can be used for the correction of results of the automated digitization of the road network. Moreover the construction of modern roadways leads to the transformation of the natural relief of the region that, as a rule, is represented by short linear objects, which describe the relief (e.g. embankments, excavations). If these objects are digitized before vectorization, then their geometrical and attribute information can be used in the algorithms to correct results of the automated digitization of the road network.

Administrative and political divisions. The borders of the administrative and political division of a territory often follow rivers, roads and other linear elements of the cartographic structure. Prior digitization of these elements can be used for the correction of results of automated digitization of the corresponding layers of raster maps.

The examples presented in this section show that the characteristics of internal structure and relationships between the vector objects can be used effectively in automated correction of errors of the automated vectorization of raster maps. In practice it means the development of more specific software for automated cartographic image recognition.

Summarizing the discussion of this section, we propose that the process of automated correction of results of automated cartographic image recognition (post-processing) follow the scheme presented in the following Table 8, where we show, as before, an expert evaluation of the degree of possible automatization of the corresponding procedures. expressed as the scores defined in Table 2.

Operation

Score

  1. Correction of vector layers based on characteristics of their internal topology

75

  • Correction of vector layers based on their spatial and logical relationships with other vector objects
  • 50

  • Final correction of vector layers in entire electronic map system
  • -

    Table 8. The degre of possible automatation of post-processing operations

     

    [ Este número | Artículo]



    Dirección General de Servicios de Cómputo Académico-UNAM
    Ciudad Universitaria, M
    éxico D.F.