For this week's lab we were given two raster datasets - they are adjacent aerial photos of the UWF (Pensacola, I assume?) campus - which had no accompanying georeferencing information, and we were asked to line them up with two already-georeferenced buildings and roads GIS layers. The accompanying map shows the two photos, which line up almost perfectly with each other (I drew a blue line across the map at the point where the two photos meet to make the division a bit more visible) and quite well with the buildings and roads layers.
Georeferencing the first raster (the one at the top) was straightforward: we just had to find individual points on the aerial photo and match them to points on the buildings or roads layers. The second raster was a bit more difficult because it had been intentionally distorted from the original shape, so after finding a few matching points we had to switch from using a 1st-order polynomial function (one click in ArcMap) to a 2nd-order polynomial, which allows the raster to bend and warp a bit in order to accommodate the georeferenced layers. I briefly tried a 3rd-order polynomial but it made things on the ground seem to fit worse, so I reverted to 2nd-order. The error stayed fairly low, but the distorted raster was resistant to the idea of lining up really perfectly. However I think it looks quite faithful to the buildings/roads layers in the .jpeg.
Apart from the relief of a lab that was less effort than Week 5's fortnight-long grind, I was thrilled to try georeferencing in ArcGIS - I tried it once last year at home with an old topo map that I had scanned in my home printer/scanner, but those results were a lot less satisfying. Evidently having the right source material, not to say the right tools, makes a difference.
No comments:
Post a Comment