GCP vs post flight QGIS for georeferencing image with a garmin handheld GPS

General Help
#1

My mosaics are far from perfect. When I use a base map (e.g. google or bing satellite), they are not only linearly wrong (like if everything were x meters away from the satellite point but always in the same direction)
But, the image is kind of warped. (sorry for my lack of technical jargon)
At one side of the image, a tree I photographed shows itself NW from the same tree in the satellite basemap, but on the other side of the image, another tree there shows itself SE from the respective tree in the satellite basemap.
I’ve previously used a Georeference function in QGIS that arranges an image. So, first, you take visually recognisable points with a GPS (I have a Garmin GPSMAP 62, GPS/GLONASS, which has average waypoint function for giving more accurate measurements. I understand this is not the ideal but it’s the best I can get.) and then you point in QGIS the visually recognisable objects (e.g. trees, a house corner, etc., and match it with the GPS points you took. QGIS does the magic and fixes the image so everything fits.

Is this a good way to proceed?
Should I use GCP instead? Some people told me GCP are only for “good, real GPS, not for my garmin hobbyist one”… Ok I understant, but If I only have the garmin, it’s better than nothing, right?

thanks!

2 Likes
#2

I’ve taken reasonable-quality GCPs with a cell-phone in a pinch, so your Garmin with an actually decent antenna and point-averaging should do just fine provided the accuracy is sufficient for your needs.

People can get wrapped up in the minutia a bit and forget that not everyone, nor every task, needs sub-centimeter accuracy RTK-surveyed GCPs (though it obviously is awesome).

I’d say that the QGIS Georeferencing tool is incredibly powerful and can certainly warp things in any way needed to match your input data. If you process with your Garmin-collected GCPs, our pipeline will work to translate, scale, and rotate the data to match them as best as possible.

2 Likes
#3

Agreed on all points. The only thing I’d add is try setting your gps-accuracy to 0.2, if this is a single flight. That should help better constrain the model. Also, keep in mind that google maps / google earth isn’t a good point of truth for accuracy, so offsets from it should be expected.

3 Likes
#4

thanks!this is a great advice… setting gps-accuracy to 0.2. Should I try to fine tune that parameter depending on something?
(yes, it is a single flight). If I make a new flight on the same area, can I combine both images into one process/mosaic? or that is not recommended?

And yes, I know satellite basemaps available on the internet are not precise, but the imprecision in my mosaics is not a linear offset, but more maybe like the bowling effect described here:
https://docs.opendronemap.org/arguments/gps-accuracy/

thanks!

1 Like
#5

Cool!
thanks!
then you would recommend first trying doing the GCPs with opendronemap over QGIS Georeferencing processing?

1 Like
#6

Yeah, I would use our GCPs first just for less touch-time with the data.

Stephen gave a great pointer, too, in that tuning gps-accuracy can be huge. Maybe set it to 2x your average gps error for the collection.

You can always combine the data, but try and make your workflow simple as you work on a new skillset.

1 Like
#7

that’s a wise advise, to keep it simple in the start. thanks.

You mean 2x the gps error I got with on the first mosaic I made with no tweaking?
And in what units are you talking ?

thanks again
I hope I am not surpassing myself with the questions and your good will.

1 Like
#8

Yeah, 2x the mean X/Y GPS error in the Report.PDF, or if you have an idea of what your mean GPS error typically is for that platform, 2x that estimate.

Nope! We’re all here to learn and help one another :slight_smile:

1 Like
#9

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.