ArcScan and various tracing methods

Blog Post created by IainS on Oct 15, 2014

I have spent a bit of time looking at this tool which applies to raster documents and offers the promise of quickly and seamlessly converting to vector polygons and lines.


There are two scenarios I have looked at


1) Conversion of historic plan and maps and I have looked at a plan of Crown Street Wollongong and the Parish of Naas plan. These are plans I had as georeferenced rasters.


2) Conversion of an archaeological site plan. I had one of a mill I excavated at Yass (NSW) which was a simple black and white plan on tracing film which I had scanned at 300dpi


These are of course related to some pressing tasks I have as a result of recent archaeological excavations I have been directing. The aim is to try and reduce the amount of time it takes to draft maps and plans.


The first stage in the process was reading the manual and watching some videos on you Tube. I then even did a Arc Scan tutorial. This was by far the most useful of the lot and printing out the tutorial gave me a workflow to follow with my test plans.


The second stage is to get the rasters into a suitable form. This involves maximising contrast, sharpening and “cleaning up”. I found this easier to do in Photoshop because of the greater control over the “eraser”. Most historical plans have a lot of “noise” that we tend to overlook. There are blots, annotations and crinkles all of which our eyes generally ignore.


In ArcGIS you have to set the raster properties as black and white for ArcScan to work which is relatively easy to do. You also have to set up snapping in the Editing tool bar and then use the ArcScan tool in an Editing session.


There are two approaches you can trace a feature or automatically vectorize. Both worked really well in the tutorial with a clean map supplied by ESRI but in my real world the results were quite mixed largely because the data was not a clean as it could be, In particular there were gaps in lines delineating buildings and portions on the historic plans and these caused problems.


You can work around this by further cleaning up or editing the resultant polygons but for something like a building or a land portion you get far too many vertices whereas if you manually traced a square you would have four rather than 20. The problem is that the raster plan may have slight variations so that a line the is straight is rastered as a bit wobbly and the program picks that up.


The archaeological site plan had to be cleaned up quite a bit but still the process got confused because a site plan is really a series of overlapping polygons and it was hard for the program to sort them out. Of course manual tracing is still an option.


Ultimately I wonder whether given the need to clean up the raster scans and then clean up the results there is that much time saved. I would normally georeference maps and plans as a background layer and then digitise polygons off them which is probably just as quick.


For plans and sections I do them in Coral Draw (or Inkscape) and then when I need to add georeferencing I do it. It takes a while and you need to make sure you have good control over layers and groups but it potentially seems to be just as fast as using ArcScan.


I have tried to use the bitmap tracing tool in both Coral and Inkscape for section drawing and found much the same problems. So it is back to the drawing board for me.