"Georeferencing" drone imagery without calculating orthomosaics?

[rocketbird]

Hi everyone,

I'm currently working on a project that involves using drones like the DJI M30T to capture imagery and create orthomosaics. Out of curiosity I'm wondering, if it is possible to "automatically" georeference a single drone image (JPG) taken with either the wide-angle or zoom camera, without going through the whole orthomosaic process or using Ground Control Points (GCPs)?
If I'm not mistaken, each drone image contains GPS coordinates in its EXIF data. Shouldn't it be possible to use this information to directly georeference the image, even if it's not as accurate as an orthomosaic, especially when using tools like RTKs?

Thx for your answers in advance.

 

[Meta4]

I'm wondering, if it is possible to "automatically" georeference a single drone image (JPG) taken with either the wide-angle or zoom camera, without going through the whole orthomosaic process or using Ground Control Points (GCPs)?

Can you explain what you are wanting?
It's not clear from your post.

 

[Kilrah]

As you said the GPS coordinates are already in the EXIF, so it's not clear what more you want.

 

[Mark_Y]

Have might want to look at the OpenAthena phone app.

 

[drhitch54]

I did not know that the exif data in a JPEG file contains more than one set of coordinates for the center of the pixels? If that data does contain the corners, then yes, it seems likely you could construct a geofence although it's not exactly on Target.

I also was told that that exif data is in lat/long not in decimal so it would have to be converted first.

Ref: How to geocode image metadata (EXIF data)

 

[anotherlab]

Hi everyone,

I'm currently working on a project that involves using drones like the DJI M30T to capture imagery and create orthomosaics. Out of curiosity I'm wondering, if it is possible to "automatically" georeference a single drone image (JPG) taken with either the wide-angle or zoom camera, without going through the whole orthomosaic process or using Ground Control Points (GCPs)?
If I'm not mistaken, each drone image contains GPS coordinates in its EXIF data. Shouldn't it be possible to use this information to directly georeference the image, even if it's not as accurate as an orthomosaic, especially when using tools like RTKs?

Thx for your answers in advance.

As already mentioned, the OpenAthena project is doing this now. You can get information about the project at Theta Informatics LLC

 

[Meta4]

I did not know that the exif data in a JPEG file contains more than one set of coordinates for the center of the pixels? If that data does contain the corners, ....

The exif info has only one set of coordinates.
Those coordinates are for the location of the drone when the image was recorded.
They only relate to the point below the drone if the camera is pointing straight down.
They have no relation to the corners and there's no way they could.

 

[anotherlab]

The exif info has only one set of coordinates.
Those coordinates are for the location of the drone when the image was recorded.
They only relate to the point below the drone if the camera is pointing straight down.
They have no relation to the corners and there's no way they could.

The location information includes altitude, in addition to latitude and longitude. OpenAthena uses that along with a Digital Elevation Model to map out items inside the images taken by the drone.

 

[AMann]

I wouldn’t call it “georeferencing” if you are relying on coordinates from a non-corrected GPS fix because it could be as much as 10 meters off. Given this, you could estimate the coordinate points - too much to describe here, but this is a good text for using photogrammetric mathematics to calculate it:

https://ethz.ch/content/dam/ethz/special-interest/baug/igp/photogrammetry-remote-sensing-dam/documents/pdf/math-of-photogrammetry.pdf

A rough way to do also this is to find the features from the corners of your photos on Google Earth and use those coordinates, but again, it’ll be just an estimate and can also have the same errors.