I learned of another cloudbased aerial photo processing service http://dronemapper.com/
I tried their free 25 image upload.
The process stalled and when I contacted them, Jon-Pierre Stoermer advised that, "...the problem with your flights is that each image doesn't have a unique lat, lon, and elevation geo-tag. I can create maps without geo-reference if you like. I'll take a look at your latest uploads and see what I can do..."
I replied, "... Makes sense, the photos were taken in continuous mode, ~1 per second. However, the gps track was from a QStarz BT-Q1300 at 5 second interval so in geotagging, same coordinates were applied to more than one image. If you can process without, I'd like to see the results. I'm pleased to see your service and wish you well. I found it through Conservation Drones connection."
After I saw the results, I wrote back, "One thing I have been doing which should avoid the gps issue you noted, is that I have been flying an iPhone with GPS-Trk app, which writes a track point every second. It can run in background while Time Fugit intervalometer runs in foreground, triggering a custom circuit to drive a pair of CHDK enabled Canon A490/495s running remote shutter (There is also a simple 555 based timer circuit to trigger the cameras). One is modified to shoot IR, the other captures visible. Ned Horning's application at https://github.com/nedhorning/RegisterPhotos co-registers near-IR and visible photo pairs. Still have to geotag the images, but should have better time resolution in the gps track than in the image sets (with remote shutter, they trigger every 2nd pulse and the minimal interval is one second).
I've just done a little testing but others have produced nice results with this combination, e. g., Chris Fastie's work: http://publiclaboratory.org/notes/cfastie/6-3-2012/nrg-images-cover-crop-trials
2 Comments
The automated output results in a nice image, or evan great.
I could analyze it critically; it does have errors, the original image sharpness is lost, there are areas where things are blended that should not be to support seamless transitions and make the image appear as one sole composite coverage when its not. And there is this whole concept of the human factor in grassrootsmapping that is partially lost when the cartography is outsourced.
But at the same time, I would consider it a good result.
I think this technique should not replace the existing manual workflow (not speaking about forever but in the "now" certainly), but can totally exist as a separate workflow for those that are either not interested in making maps by hand, or would like to outsource the job to software for any number of reasons. It seems like the folks at conservation drones are only looking at this solution.
It would be fun to have a Kasparov-esque people vs. machines showdown. But I remain steadfast on the value of the human element in making these maps. I just don't see it as a nut that is yet to be cracked.
I hope I do not sound discouraging. I just propose that people consider this methodology as separate today (who knows how people like me will evaluate this in the future). There is something very good there, no doubt. I just am not ready to see it as the inevitable "goal" of what we are working towards. And I do think that you are making a real contribution for others as you work towards finding an automated mapping solution and document the efforts as you have done here.
Thanks Pat! -Stewart
Reply to this comment...
Log in to comment
Thanks for the comment on the note. It provides a terrific perspective on the difference between personally made maps vs the automatically generated ones. Very thoughtful and balanced. The automated approaches really seem to be coming up with more folks flying autonomous drone missions and wanting the processing of images to be automated as well. As an example of the difference in approach, I still recall your pointing out the choice to include a view of graffiti on a wall on the Gowanus map when you were explaining your PhotoShop workflow.
Is this a question? Click here to post it to the Questions page.
Reply to this comment...
Log in to comment
Login to comment.