Picture geotag information

[Joe Paulin]

JField geotags pictures taken with the LS... I am interested in knowing more about this. I assume it uses the spot position of the LS at the moment of snapping the picture? Is this correct or does it use a less precise position? I have collected a few survey points using VRS RTK and had the LS set up to take a picture at the start of the observation. I set my project coordinate system to WGS84(ITRF2008) with NAVD88. When I inverse between my surveyed coordinates and the geotagged coordinates of the pictures, I get roughly a foot horizontal distance and the heights of the pictures are roughly 110 lower than the surveyed coordinates, not taking into account rod height. Are the geotaged heights ellipsiod heights? What latitude and longitude shows up on the geotag?

I see a great value in geotagging pictures taken with the LS with the exact survey grade position of the LS at time of picture capture...

Thanks for looking into this, and by the way, what an awesome software release/update!

Keep up the great work.

 

[Mikhail Drakin]

Pictures are geotagged in plain WGS84 (ellipsoid height); coordinates are taken from the most recent epoch processed by the device at the moment of writing the file, be it RTK or standalone. We have always considered geotagging as a tool to easily drop photos in a tool like Global Mapper for illustration purposes, so standalone position should be enough. Please describe the application you see for photos with exact position.
Photos are never modified after been taken, and as current workflow ideology suggests postprocessing in many cases - even if we write exact geotags, they will most probably diverge from source point later.
Please also note that standard EXIF format implies using WGS84, so in case you request geotags in some other coordinate system, you will most probably need to write your own software to interpret them.

 

[Joe Paulin]

Good morning Mikhail, thanks for your reply. One of my applications for the LS is using pictures taken with the LS for photogrammetry purposes - to supplement aerial nadir photos to create more accurate 3D models of buildings and items with vertical relief. With the LS, I can photograph low level building profiles much easier and in much more detail than with a UAS, with survey grade geotags. With this purpose, the more accurate the geotag, the better the final result is. With this in mind, would it be possible to geotag the photos with the orthometric height instead of the ellipsiod height? That would streamline the process quite a bit. I think the Lat and Long is fine the way it is.

Also, I have a related request/need to be able to export text/csv files of survey points with a little more data than what currently is available.

Can we add a few more data fields to the export options? What I am looking for is an azimuth of the LS at picture capture (taken from compass reading) and tilt values (roll, pitch).

The text file would look like this:

Point name, Latitude, Longitude, Orthometric height, Azimuth, Pitch, Roll

Please let me know your thoughts, this is a very versatile tool.

 

[Mikhail Drakin]

Joe,
Photo acquisition and point acquisition currently are separate processes, so in many cases data contained in them may diverge (as I've mentioned, photos are not touched after been taken, and points are likely to be postprocessed). Moreover, photos may be taken before the respective point has been calculated. So if we implemented what you are asking about, for most users this would produce inconsistent data. Moreover, you would also most likely sometimes get incorrect data without noticing it (because it would be only sometimes). Consistent point coordinates are contained only in the project database, not in JPEG files.
Both your tasks (getting coordinates and getting attitude) can be easily achieved via scripting. Any photogrammetry software I am aware of accepts exterior orientation data in CSV, which overrides geotags even if they (geotags) are present. So I suggest that you once create a script which will produce necessary EO file from J-Field CSV export and JPEG image files (to link image file names to point names, use "Media / Image Files" CSV field), and just launch it each time before using LS images in processing. In order to extract EXIF from photos, I recommend using ExifTool (http://www.sno.phy.queensu.ca/~phil/exiftool/).

 

[Joe Paulin]

Mikhail,

Given your reasoning above, I agree with your stance that things should be left "as is" in regards to how the photos are geotagged. I use Agisoft Photoscan, which does have the ability to import EO data in CSV and I think that creating a csv file is the way to go. That being said, I was unable to export a CSV file from JField with the "Media / Image Files" CSV field. The "Image Files" field was empty in the created csv file. Also, I see that the camera azimuth is displayed in the exif data but I don't see any tilt values.
How difficult would it be to add receiver azimuth and tilt information as values that can be exported in a surveyed point csv file?

Thanks for looking into this!

 

[Mikhail Drakin]

Joe,
Do you have image files attached to the points (i.e. these images must be listed among the point media in Points screen)? If they really are, and image files for these points are not listed in exported CSV, please either post the project archive (with images included) here, or use "Send to support" and put your receiver s/n in this thread so I can investigate.

As for the attitude data, I don't know any standard EXIF tags for pitch and roll values. Strictly speaking, even the "direction" tag I write now is for speed vector, not camera orientation.
However, I write full attitude to the Maker Notes tag in XML form like :
<?xml version="1.0" encoding="UTF-8" ?><JavadGNSS><Inclinometer Rotation="204.1513000504558" Alpha="0.10543809758893744" Beta="0.14374685449287311" /></JavadGNSS>
Here Rotation is azimuth, Alpha is pitch and Beta is roll (all in degrees).

 

[Joe Paulin]

Mikhail,

I just sent the projects to support, s/n 397. It is just a simple project for testing, to summarize, I collected a few survey points and attached photos to the points. This was done before I updated to JField 2.0, thanks for looking into this.

 

[Mikhail Drakin]

Joe,
There indeed is a bug in export. Thanks for finding the issue. I'll see if the fix can be included in the nearest release. In the meantime, add "Raw files" field to CSV export also, it will force including image files list (yes, it's a bug that silly).