Allowing 2-GCP georeferencing: how should we do this?

[mradamcox]

Recently we've completed the implementation of the Helmert transformation which allows to you georeference a layer with only 2 GCPs. While this is extremely useful in many contexts, it could also be abused pretty easily I think, leading people to create badly fitting georeferenced layers without even realizing it.

Currently, the following behavior is available to a handful of users with certain permissions (let me know if you want to try this out but don't have access):

1. Create two GCPs
2. In the transformation settings, the Helmert 4-param option is now available, and you can switch to it manually
   
3. A preview appears like normal, based on the 2 GCPs
4. If you submit your work, this warning appears:
   
5. Clicking through will submit and warps the image like usual
6. If you instead go back and add a 3rd GCP, the interface switches automatically to Polynomial, like the existing default has always been.

I'd love to hear other ideas or preferences for how this behavior could/should be improved! I think it would be great to make this more accessible to more people, but I'm not sure the best way to do so.

---

As the georeferencing interface gets more advanced, and has more complex capabilities, I think it's appropriate to further "curate" the experience in order to guide people toward best practices and the creation of high-quality work. This is especially possible because we are focusing on a specific set of maps with a specific set of characteristics... In the future, with different content, it may be appropriate to have different behavior/defaults in the georeferencing interface.

[mhpob]

I think this is an incredibly worthwhile addition -- thanks for adding! I do think this actually has less of a chance of being abused than TPS, which requires a ton of GCPs to offset the warping and can be made to do whatever you want. At least this is only "spinning" the map.

GCPs are not created equal and depend on the method and the scan you're dealing with. As OIM is a crowd sourcing platform, individual error is to be expected. Perhaps the method used can be reported when perusing the georeferenced pages to allow people to run a high-level filter on how "trustworthy" they feel the page is?

[mradamcox]

I like the idea of better pushing information about how a layer was georeferenced into various interfaces that display or list it. Just the basics like number of GCPs and what transformation is used would be a start. Ideally, this would also be identifiable from the preview and other map interfaces.

Good point about TPS, it's just as easy to make bad fits with that transformation, maybe easier. Expanding what I mentioned above, we could consider the idea of making georeferencing presets or profiles (maybe there is a better term). Basically, considering these factors: transformation options, "target projection" (which is disabled but is in the system), available reference layers, minimum allowed GCPs, etc., a preset/profile would disable certain options, thereby making the behavior of the interface ideal for certain types of maps.

For example, right now, the whole site just has Sanborn maps (basically), so the preset/profile could look like this:

• Polynomial and Helmert, no TPS
  • Maybe allow TPS for Key maps?
• No target projection switcher
• Key maps as reference layers
• Parcel layers available (if in system)

On the other hand, if you were georeferencing small scale maps of entire states or countries, you would want a different preset/profile, like this:

• Allow all transformations
• Enable target projection switcher
• No reference layers

Anyway, food for thought...