Molnár Gábor – Timár Gábor – Biszak Előd: Can the First Military Survey maps of the Habsburg Empire (1763-1790) be georeferenced by an accuracy of 200 meters? Conference: 9th International Workshop on Digital Approaches to Cartographic Heritage, Budapest, 2014. Volume: 9. 127–132.
9th International Workshop on Digital Approaches to Cartographic Heritage Budapest, 4-5 September 2014 In our earlier version of the First Military Survey georeference, we selected several tens of control points per a region of the Empire. Using the image coordinates and the UTM coordinates of these points we estimated the 12 parameters of the best-fitting quadratic transformation between the coordinate system of the scanned image and the UTM grid coordinates. This transformation was applied to the whole virtual mosaic of the sheets covering the given region. The accuracy of the horizontal control was quite low: from several hundred meters (in case of smaller provinces) to even almost two kilometers (in the extremities of the larger ones). However, this solution was the first working one for this nice piece of cartography (Tímár et al., 2011). The technology of the correction grids (grid shift binary: GSB, aka. NTv2; for details, see Tímár et al. B, this volume) offers a remarkable tool for refining the georeference of historic maps (Molnár & Tímár, 2011). In this method, we use again ground control points (GCPs), identified terrain objects that can be found in the old map (giving their image pixel coordinates) and also in modern cartographic products (providing their location in e.g. WGS84). In the present study, we show the difficult example of the First Military Survey, where there is no initial information about the native coordinate system of the map sheets. Data and method At the time of submitting this paper, processing of map sheets of two Habsburg provinces (Lower and Upper Austria) have been finished and published, another one (Bohemia) is ready, while processing of Croatia and Galicia have been started. For all of these provinces, a 'province image mosaic' was made, using the scanned images of the map sheets. We collected several GCPs (minimum 3 per sheet, usually identified town or village center objects) in these mosaics. For all GCPs, we record two coordinate pairs: 1. Their pixel position in a 'province image mosaic', and 2. Their geographic (WGS84) coordinates, according to Google Earth. As we had no hints for the initial coordinate system of the maps, we used a classic engineering method: define one somehow and provide a correction procedure that refines it to acceptable level. The input dataset for both steps was the GCP set. In the first step, we estimated the parameters of the best-fit Cassini projection, providing the connection between the image and the geographic coordinates of the GCPs. The estimated parameters were the following: • Cassini-grid easting and northing of the upper left (NW) corner of the province image mosaic • E-W and N-S extents of an image pixel of the province image mosaic • longitude of the best-fit projection origin • Cassini-grid easting and northing of the best-fit projection origin One can immediately observe that the latitude of the best-fit projection origin was not estimated. It was not indeed, as this estimation is invariant to this value. In case of a Cassini-projected map, the latitude of projection center can be set to anywhere, if the false northings (the northing position of the projection center) are altered together with it. In the present procedure, we provided a pre-selected latitude for the projection center, and during the estimation, we get the northings of it (the last element of the above list) according to this value. Using these figures, we can geo-refer the province mosaic by a single linear fit to a Cassini grid with the estimated projection, interpreted on the WGS84 base surface. The horizontal accuracy of this fitting is the same we had earlier by the quadratic fit to the UTM coordinate system. Interestingly enough, for Lower Austria, the estimated central meridian (longitude of the projec
