Coordinate system/projection issues

Hi guys,

I have recently uploaded a couple of projects with geotagged photos as a means of georeferencing rather than my standard process of using GCPs. The resulting maps seem to be accurate, except showing at about 80% of their actual size. I believe this is a projection system issue, can anyone help us solve it?

We assumed Pixhawk tags images using the WGS84 coordinate system, so checked resulting maps against WGS84 sources such as Google Maps and Bing maps etc., but all are showing our map as slightly smaller than the imagery it is overlaid upon.

This issue is not limited to DroneDeploy, we have tried other services with same outcome, but hope to go forward using DD so will ask the question here.

Thanks in advance

Here is on of the sites with the problem, you can see some landmarks in the background that are bigger than shown in the map https://www.dronedeploy.com/app/data?planId=5601db7931ff9a74bcaddadb

I’m wondering if it is an AGM / ASML issue but can you post more about your setup, particularly pixhawk fw, mission planner / GCS version, camera and lens? Given it affects multiple platforms have you asked over on diydrones? Also, as it may be relevant, where in he world are you?

Hi @agis, @rogerpearson is asking exactly the right questions, and makes a good point. I’ve checked the EXIF logs for the plan, and I see you’re running a GoPro HERO4 Black. What did you use to geotag the images?

Am I right in saying you’re only a few miles off the coast in NZ?

Hi @rogerpearson and @jono, thanks for your replies.

Firstly , Roger, set up in the map listed was 3DR Iris plus with Gopro Hero black as Jono discovered from the exif data. We geotagged the images using Mission Planner. This was a while ago, but would have been the latest stable version of MP in May. Also unsure the PH FW for that flight sorry, but latest flight was performed using 3.3.1 I believe.

@jono, we are in New Zealand, but not off the coast. In the map linked you can see the site is in suburban Auckland, so I wonder where you got the information suggesting we were off the coast? This would be consistent with a projection issue as suspected, although it does seem unlikely to be the cause.

I have since processed another test flight, this time using an X8 with a sony a5100 with pancake 16mm lens. The results are even worse, with not only the scale being off (this time too big!), but also the location of the flight is off by 50m or so, most of it didn’t process, and what did process looks unusual.

I geotagged these images using Event38’s image tag utility, as MP’s function wasn’t working for us.

To give a bit more information, we have been professionally flying UAVs for about 6 months, and have a workflow that performs very well that does not involve geotagging images, because of issues we have experienced with doing so. Services such as DD are very attractive to us, so we must revisit out workflow to suit, and are running in to these issues again!

Any help much appreciated.

Cheers guys

And here is the new flight I forgot to include https://www.dronedeploy.com/app/data?planId=56244bd431ff9a04a7b84c7c

Hi @agis, thanks for the info.

Regarding the ‘just off the coast’, I guess what I mean is that you’re fairly close to sea-level. Placing the pin from you data says it’s about 1.5 miles from a large body of water at sea level:

Regarding the second set of images. The reason it didn’t stitch is because you’ll notice that the center of your image is focused, but the edges (used to stitch) is very unfocused:
Center:

Edge:

This is due to the very low F Number on your camera. If you notice, it’s set to F2.8 (the widest possible aperture for that lens), which means you will have a very shallow depth of field ([one randomly selected tutorial][1]).

Try increasing your aperture to F8 or above, and you’ll see much crisper edges, and hence much better stitches. To compensate for the change in exposure, you’ll have to slightly increase your shutter speed and ISO.

The GPS tagging is still a mystery. Do you still happen to have the logs for the flight? If you could send those through to support@dronedeploy.com I’ll take a look to see if there’s anything we can do. This might take a while to get to given current backlog, but I’m interested in what’s going on.
[1]: http://www.cambridgeincolour.com/tutorials/depth-of-field.htm

Hi Jono,

Thanks again for your help, I understand what you mean about being near the coast, I thought you had seen a geotag that placed us in the ocean near NZ!

Yes the aperture was too wide, thanks for pointing that out, we will adjust our method going forward. We have uploaded this same dataset to other online processors, and they have stitched the images no problem, other than the geotagging issues. Here is the output from one such service

Yes we still have the logs,and the kml looks good, but as you say the backlog will probably be too long so we will take the discussion over to diydrones or similar!

Thanks for your help, we are perplexed by the issue which is why we haven’t been using geotags at all over the last 6 months. It seems all online processors want us to use them, so we had better learn.

Can you suggest a geotagging utility other than mission planners built in tool?

Cheers

Thanks for letting me know about that - I’ll get one of our engineers to dig a little deeper.

The reason I’m asking for the logs is that I’d love to automatically ingest these logs and do the geo-tagging for you. For that, we’d need a few example sets of data, so even if we aren’t able to correct this issue for you immediately, it’d be great to have the info :smile:

With that in mind, I’d appreciate it if you wouldn’t mind sending them across to support@dronedeploy.com

I pulled your dataset off DIYDrones and had a quick look this afternoon…

Location in DD looks OK. I’ll post a link when it’s finished processing. Can’t zoom further with this map / location but it looks good from that point…

Was that a manual flight? Flight path, camera angle and time / position between shots are all irregular.

This was tagged in MP…

I know that’s not a lot of help to you, but so far I’m not seeing the issues you are seeing with your data!?

Finished map here

Hi Roger,

Thank you so much for taking the time to look in to this, we really appreciate it.

Firstly, no the flight was not manual, it was a grid survey created in Mission Planner, set to trigger at a set distance. The camera was not on a gimbal, so the angles will change depending on the direction the UAV was flying. I suspect the flight path irregularities are due to either wind, of which there was some (10-15 knots) Or waypoint radius. I have since checked and waypoint radius was set to 3m, so shouldnt be a problem

The map you have produced on DD looks pretty good, about what we would have hoped for. What is it that you have done differently to us to get that result. As I understand it, we both uploaded identical imagery and logs, yet received vastly different results!

Here is our resulting map on DD for reference https://www.dronedeploy.com/app/data?planId=56244bd431ff9a04a7b84c7c

We flew a large site yesterday and ran in to yet more .log issues, this time getting z data of 10’s of 1000’s meters high, when we flew 100m or so above sea level! the logs were also 200mb, which is bigger than anything else we’ve ever had by a factor of 10. At this stage we are stuck using a stitchign service that lets us georeference from a basemap.

Thanks again for your help

Reuben