Discussion: Updating Maps???

Our club has a small library of older maps that were developed in OCAD but prior to wide use of GEO referencing. Is anyone aware of a relatively easy process to geo reference older, existing maps? Or are we looking at a complete re-draw?

Thanks

Try this way.
Open the map file.
Open a georeferenced photo ( like from Google Earth in background) or an open street map of the area.
Then instead of moving the photo to the map you move the map to the photo.
Do that by selecting Map from the upper toolbar
>Transform from the dropdown menu
> Affine from the second drop down menu
Then select/ click on area on map and then matching area on the photo in background (just like adjusting a background map but only in reverse.)
Do that twice followed by ENTER and the map should move to the georreferenced spot.
I THINK that will work. I have only had to do it once.

Howard, it really depends on what you are looking for. The process that Gord describes can work and in certain cases can even work reasonably well. Namely old forest maps with minimal man made objects. It doesn't work well for sprint maps where you really introduce noticeable distortions to man made shapes. For forest park maps you end up with a mixed result.

Of course, if you have an old pre-georeferenced map it is probably in need of significant updates anyway and is certainly not from a lidar base so will have an older feel. If you have the money and a map that is used annually it would be worth looking at starting from scratch.

If you want to discuss more send me a note at jeff@navigationsports.ca and I can take a closer look at what you've got.

Edit - I only quickly read and misunderstood what Gord had said.

Yes Gord's solution will add georeference to the map but it doesn't do anything about distortions which these older maps always have. What I thought he was describing was a process whereby you can stretch and warp the map to remove distortions but it's far from a a perfect process and in sprint maps you end up shift building to the right place but skewing them so they're no longer rectangular.

In OCAD 2018 (I don’t know about older versions), you can use the tool rubbersheet. This allows you to load an aerial image as a background image and then assign points that can stretch / skew / assign known locations in order to georeference. I have found that generally it’s likely not worth it, unless the original map was done really well and matches somewhat closely, the odds of not having warp are low. But depending on that and the use you’re going for it can get you by. But essentially the recommendation is to redo with Lidar so you have a proper map for many years to come.

Note that an aerial photo even if georeferenced may include as many distortions as the original map. Unless the photo has been properly and well ortho-rectified against a reliable DEM.

If you're going to go down the rubbersheeting road (which generally speaking I wouldnt recommend) make sure your rubbersheeting to lidar data and not photography.

OUSA has a nice set of slides on how to do this (linked from Resources > Mapping):
https://docs.google.com/presentation/d/1Po1HuwcuLK...

I think this was put together by Jon Campbell, but there is no credit listed, so not 100% sure.

@smittyo - that's Jordan's

My take on this is that if you are very fortunate in the quality of the old maps, you might get very lucky and be able to achieve pretty good georeferencing without much effort. If you're somewhat less fortunate, you might be able to achieve pretty good georeferencing with a fair bit of effort. If you aren't fortunate, no amount of effort is likely to be sufficient to achieve pretty good georeferencing throughout the mapped terrain - some older orienteering maps contain spatial distortions varying wildly from one part of the map to another and you'll drive yourself crazy trying to adjust it so one area is accurately georeferenced without screwing up other areas. Even if you can establish the true location of every point feature on the map so you can do a rubbersheet transformation of the old map to pull all of them into position, it's unlikely the original field checking located everything with perfect relative accuracy so small distortions will remain. And at that level of effort, if LiDAR is available, you might as well start over from a new basemap and have everything as undistorted as possible.

That said, georeferencing old maps has been done successfully by people who weren't crazy and apparently didn't regret doing so. My understanding from personal communication is that the map of Vasquez Rocks used for the Middle at LAOC's Nationals last November was produced by painstakingly aligning the older map with a new LiDAR base and then transferring/redrawing the old fieldchecked features (plus spending some time in the terrain afterwards updating the vegetation mapping and anything else that had changed in the terrain).

In my personal experience, in my first professional mapping project back in 2016, I was handed a non-georeferenced photogrammetric basemap that had been procured roughly a decade earlier and then sat on a shelf waiting to be used except for a small section that was fieldchecked and used for a training camp. There was no LiDAR coverage of the terrain at the time so I georeferenced it more or less as recommended in that OUSA-hosted slideshow, except that instead of using LiDAR or orthorectified aerial photos as a reference, I went out into the terrain and recorded GPS waypoints on a few dozen identifiable point features scattered across the basemap (mostly dot knolls, pits/small depressions, and sizable cairns) and did a rubbersheet transformation of the basemap to pull all those point features to the GPS-recorded locations (plus making sure a few major road junctions within and around the edges of the terrain were where aerial photos said they should be). I was lucky - that worked well enough I was able to use GPS surveying techniques almost everywhere in the terrain after that without running into noticeable issues where features I placed by GPS clearly didn't align with the contours. "almost everywhere" since even so I ended up doing another rubbersheet transformation along the NW edge of the mapped terrain, where a mismatch was obvious on the road along that edge of the park and the terrain for several tens of meters heading into the park.

The best bet is to use as reference points features that were not put in by the original mapper. That is to say, the mapper may not have placed something like a boulder exactly (either inadvertently, or on purpose for reasons of legibility), but something like a road intersection or a building that appeared in the stereo photos should be in the right spot, and make a good anchor point. Trying to fiddle too much with rubbersheeting is likely a process that's not worth the trouble. If I were doing this, I'd be looking for three solid reference points that are separated as far apart as possible, do the transformation, and then see if it looks close enough. Running a major trail through the middle of the map to generate a GPS track, and then pulling that into the map file, is a good test. If it's clearly off, I'd start over. If it looks reasonable, I'd stop there.

As an additional note, the hillier the map in question, the tougher it's going to be to try to rubber sheet it and get good results.

I will hazard to say that if the aim is to produce a high quality map--and especially when it's covering a larger area--then most folks will usually be better off to create a new base from lidar, and use the old map to identify and/or transfer features into the lidar base, and to use as well when surveying in the field with the new base.

The original question was how can he geo-reference an older map. So that is what I answered.
Suddenly you are all getting the impresssion that he wants to create a new map while using the old information. If that is the case then what Swampfox recommends is a very valid way to go. I have used it several times as an effective, time and money saving method. Create a new map and put the old map in background, trace in details then check them in the field. Yes, a cliff or boulder may not be exactly in the correct spot but at least the fieldchecker knows that there is a cliff/ boulder there to check. Saves TONS of time!

True enough, but then what is the point of bothering to georeference in the first place, if not to correct distortions and scale issues in the original map, and/or to prepare it so that various georeferenced products can be used to improve the map? So it still becomes a question of time investment versus desired end product quality.

And if you're happy with the map as is, why bother to georeference it? Just keep on using it until you're not happy with it anymore.

Our club has a ton of city maps created from ortho sources that aren't distorted but are from a time when saving a Georeferenced version was not really a thing. For these, ocad is really useful for quickly adding a Georeferenced osm or sattelite basemap and then doing a quick affine, but I would echo that if it is not a 100% match then rubbersheeting is probably not worth it due to the distorted results.

Once the map is Georeferenced, using a Strava heat map as a background map/template is awesome for confirming/correcting the accuracy of trail locations or adding in newer trails.

Swampfox I'll give you one more reason to have geo-referenced maps. When you hold a competition many of your participants are going to want to have Livelox or Route Gadget available to upload their routes.
Sadly neither Quick Route nor the good old red Sharpie does it for them anymore. No georeferenced map: no Route Gadget.

"No georeferenced map: no Route Gadget."

What?! That is simply not true.
For anyone who needs eividence, there's plenty of examples at https://www.dvoa.org/gadget/index.php

How does that work Eric?
I can't get a gpx track to land in place if the map is not geo-referenced.
Of course one can manually load tracks but are you saying you get gpx tracks into Route Gadget on non-geo-referenced maps?
How does that work?

Swampfox's "keep using" option was not one of the options at the top of the thread, but it certainly should be.
NEOOCofficer, have you answered the question "why"? I hope the answer is better than "keeping up with the Jonses".

There are some valid reasons, but there are plenty of non georeferenced maps that are still fine, even for national events.
My mis-read NEOC probably has some, I don't know about NEOOC.

The key is the quality of the original fieldwork. If the map has good relative (feature to feature) accuracy, then use it.
These relative inaccuracies are the only distortions that can be perceived by the orienteer or course setter, not the absolute distortions that might take place over multi kilometers.

Furthermore, a georeferenced map might do little to nothing to address the critical relative distortions which occur at the fieldwork level.
A state of the art Lidar base will help more, but lousy fieldwork and lousy GPS applications can still be done on a great base map. I've seen plenty of examples.
Likewise good fieldwork and GPS applications can be done as well.
The key here is knowing your personnel.

Gord, you are asking the wrong guy. I have absolutely no idea :-)
I'm only an observer.

@gord, tracks need to be manually adjusted in Route Gadget if the map isn't georeferenced (the old-fashioned way).

And if you're happy with the map as is, why bother to georeference it? Just keep on using it until you're not happy with it anymore.

It is all about changing your perception, learning to be happy with what you've got,
---learn to stop worrying and love the bomb

QuickRoute works with non-geo ref maps by letting you pull and tug on the map until it lines up with the GPX track. RouteGadget does something similar to get the map image aligned with the GPX track. In RouteGadget this done by the web admin, not the runner.

And Livelox has a tool to roughly georeference the map when you upload it, so you can use any (reasonably non-warped) map there.

Old versions of RouteGadget used split times to auto align gps track on course. Alternatively user could choose a manual mode without split time use. In both modes you could align gps track like in quickroute (with mouse right click you add/remove blue dots and then drag them to move the track on map). If admin had added 3 reference points the map gets sort of geo-referenced and in tracks got aligned right away on manual mode.

Current version of RouteGadget is entirely different and all new. The map image needs to be geo-refereneced and gps tracks can be modified (typically you just drag gpx track a little to correct those +-20 meter gps or map errors). If map is geo-referenced you export png + pgw or jpg + jgw. If the map is not geo-referenced you will have to geo-reference the image in RG (you drag drop it there and align it with background maps or gps tracks by dragging and rotating it (your map image gets 50% transparent) using corner handles. I personally use this a lot (for my scanned race/training maps so I made it really fast and handy for myself). The 3D mode of the current RG version uses the same geo-referencing and the webp animation tool too. Note, current RG can also be used without installing anything or event organizer even knowing about it, just open maps and gpx tracks and animate/compare without uploading anything to internet.

So map does not need to be geo-referenced in Ocad for RG. But map not being badly distorted is an advantage. I typically almost always manually geo-reference images in RG, even if map is geo-referenced, because I prefer editing map images in a photo editor (crop/resize) and also can visually see referencing is correct and fix it if needed.

Continuing the slightly off-topic discussion, I do the RG for our club and gave up years ago trying to consistently get the georeferenced maps from each course designer, especially as they will sometimes be using different tools. All I ask for is a PDF of the blank map and one with all the controls, along with the IOFXML course file - the same things the SI team need. Although I may have access to the source file so I can get the geofencing myself,I generally convert the PDF to an image, shrink to a reasonable size then use https://map.routegadget.net/ to stretch and adjust the image and produce the clipboard name of the map that contains the corner locations. I have my own app that handles the shrinking and takes in the new name and produces the paired coords list of the four corners (3 used for RG). In general the process takes 5 to 10 minutes assuming the original image was well georeferenced originally. For the purpose of RG this is good enough. it allows users to drop a GPX file and not often need to adjust it (and keeping in mind the limits of the GPS accuracy depending on the device and other situations). In cases where I can use the mapping tool I tend to produce a KMZ and pull the bounding corners out from that. I'll note that we are using the original RG (we do not host it).

Rough and dirty georeferencing method.

Pick an easy feature near the middle of the map. Go to the feature with a GPS receiver. I use a Garmin Gpsmap65 which uses three different GNSS satellite systems to deliver 1.5 meter accuracy in open areas, which is within the normal accuracy tolerance of O maps. Any modern GPS handheld will do. Record the Lat and Long of your chosen feature in the digital format. Load the old OCAD map into OOMapper and from the toolbar select Map > Geoeferencing and check that the coordinate reference system selected is the same as you used in the GPS receiver; probably either UTM or EPSG3857. Pick your chosen feature on the map by clicking on the obvious button. Enter the lat and long in the relevant box. You can update the magnetic declination on this page as well if that needs doing. Click OK. Save the map as either .omap or .ocd, as required.

The map is now georeferenced within the Cosmic scale of things. The chosen feature will be spot on, but the accuracy will drift off a bit the further you are from the chosen feature. That should be good enough for anything except a major event, in which case you might want to draw a brand new map. Personally I think that the magnetic deviation would be my priority on an old map rather than the "rubber sheeting",

Agreed on the priority of north lines.

One reason for georeferencing old maps is so that they can be used in apps like UsynligO.

I may be wrong, but I thought a map does not need to be georeferenced to be use UsynligO. IOF-XML course file needs to be, but not the map, map is just a pdf. It might be possible to create empty georeferenced map, import gps tracks visiting controls, then draw courses using gps tracks and export courses IOF XML. Then draw same courses on ungeoreferenced ocad map and export pdf course files. Like this big dirtortions of the old map does not matter. But I have never tried UsynligO myself so I may guess it wrong here.

Yes, just the xml file needs to be georeferenced. But it seems easier to georeference the map (as explained at the top of the thread) if it isn't already than to do anything else.

Further to my last, if you don't have the spare time to actually visit the center of every old map with a GPS receiver, and the chosen relevant map feature is visible on Google Earth, you can use the https://www.latlong.net/ website to zoom in on the spot, right click and copy the Lat/Long in the correct format. For instance, the website marks the entrance to the car park at Fort Ancient, Ohio at 39.407059151018125, -84.08965163447083 . That is on the Google preferred coordinate reference system of EPSG3857.

If you happen to be in that part of the world it is quite an interesting place to visit.

Coordinates with eight digits after the decimal point imply an accuracy of one millimeter.

A precision of one millimeter, not an accuracy of one millimeter. ;-)

Yep, despite the inherent vageries of Google Earth pictures, accuarcy is hopefully better than 1 meter, which is probaby enough for pumping life into old maps. And it's faster than drawing a new map. It just depends on how annoyed the competitors get with the end result. Everybody's a critic.

...Google Earth pictures, accuarcy is hopefully better than 1 meter.

i think there are some boundaries to your experience Tinnishill. They can be quite bad round here, with height distortion present. And between imagery dates, points jump around. I have a feeling that the default imagery may be chosen for its picture quality rather than its up-to-dateness.

I think gordhun's method will change the way map looks, map will be rotated, stirpes will not stay in place and so on. Better set georeferencing parameters right without touching the map.

Tinnishill's method uses just one point, so unless scale happens to be just right and and rotation guessed correctly there would be easily too big error for overlaying gps tracks. But if you measure an other point and use it to calculate(or iterate) rotation(declination) it will get good enough with the map looking exactly the same (because only georeferencing parameters are changed, not map itself). Sometimes you will have to also edit scale factors (or scale the map, but that will again change the way it looks).

Are you getting all this Howard?
You may be finding a lot of guys who do not care for the KISS principle.
I'm curious about what outcome you are seeking and if my original suggestion will work.
If you care to, send me an OCAD file of one of the areas you have in mind. I will try the Affine method and send it back. You will either have a geo-referenced file or you will not. e-mail gordhun at rogers dot com
And, once reporrted, we will all know whether it works or not.

I have used both rubbersheeting and affine transforms to be able to use old maps as a background layer for a new lidar-based base map:
For JWOC2015 the old (paper-only) map of the long distance area turned out to be amazingly well constructed (Helgesen of course!), so that a simple affine adjustment was sufficient to make everything line up with a few meters. I.e. all point details on the old map could then be located by the surveyor and placed exactly.

For the middle distance terrain the old map was OCAD, but based off badly ortho-corrected photogrammetry, so that I had to use several iterations before I got something which managed to align both the roads and houses at low elevations and the top of hills which were skewed by about 30 m.

The key is that the new map has to be completely redrawn, the old/warped/adjusted map is just a very nice background resource.

I am the mapping coordinator for NEOOC. All the maps I’ve created are automatically georeferenced because both the LiDAR and air photos used to create the base map were trustworthy. The maps Howard refers to are from the 80s and early 90s. They were created by foreign mappers using USGS maps as a starting point. I am told that these mappers relied on compass and pacing to locate features. I have georeferenced these maps with various versions of OCAD. The roads surrounding the venues are perfect, but nothing in these boundaries can be trusted. To join the seven maps together to create the Python rogaine map, I did use rubbersheeting on the older maps to make their margins mesh with the newer georeferenced parts. Attempting to correct all the interior map details with rubbersheeting on these old maps is just too much work, especially since I can find my with these older map without trouble. The complaints arise because of the need to have the course maps play well with course review apps. I find it much easier to start over with the newest LiDAR and air photos while using the latest ISOM symbol set. Our club is beginning an experiment using a new basemap. Club members will use the new OCAD Sketch app to add details to the map – one small section at a time. Because every mapper has their own style, time will tell if this approach can work. I hope it does. All orienteers can benefit from the process as they thoughtfully immerse themselves in the landscape as they try to capture it in line and color.
Howard really started something! I’ve picked up a bunch of great ideas from the lot of you. Even Yurets!

Our club had a few old-school maps that we georeferenced mainly to make it easier to use Livelox. An additional benefit has been the ability to assist course planners who can now load their "all controls" maps into Avenza or a similar app, and either confirm their control placement, or if the original fieldwork was less than perfect select alternative control features. We don't allow course planners to alter the map without reference to the original mapper (or an experienced mapper if the original mapper is unavailable).

All our maps have been converted to OCAD. Our technique is to add registration marks to the four corners of the map, export a TIFF version of the map, then use an orthophoto in QGIS to georeference the TIFF image, bring the georeferenced image into OCAD and use the affine transform.

This works reasonably well as long as the original photogrammetry and fieldwork are accurate.

I think its time for a show of amusing magnetic north lines after a rubbersheeting.