-1

For my masterthesis, I'm working on a GPS-independent localization system. In order to test how well it works, I want to take images in the city, however I need the "ground truth" x, y and z coordinates of the camera location.

However I don't have budget for e.g. an RTK GPS solution. I was hoping someone here has an idea on how to recover my camera location in a accurate-ish manner (around 20cm error would be ok).

P.S: One idea I had was to recover the location using Google Earth Pro, but I can't get the accurate elevation. I'm looking for something going in this direction.

Improve this question
2
  • 1
    "The City" is a difficult environment for GNSS, with signal multipath the main problem but satellite masking also contributing. GNSS accuracy of 20cm or better in an urban environment is not easy to achieve, even with Survey-grade RTK GNSS receiver. Another problem is that you're likely to need to have camera direction, and anything using a magnetic sensor is likely to behave badly in a built-up environment. Sorry, just problems, not solutions.
    – Trams
    Commented Feb 6 at 21:14
  • 1
    Did you discuss the problem with your academic supervisor? It's probably easier to find an in-house solution at your university, than on the internet. For example, they may have physical reference points in the campus with known coordinates (mine had them). Or they could connect you to municipial/state/private company surveyors so that you can use their ref. points. Or maybe there some other student who uses RTK or TS on their project, and you could just tag along and snap pictures wherever they measure some points, etc.
    – crafter6029
    Commented Feb 7 at 7:43

1 Answer 1

Reset to default
1

You can't with over-the-counter GNSS enabled cameras. Maybe your university has a PPK GNSS solution for your use? Take your pictures while the GNSS unit is running in the area of interest. Differentially correct those GNSS data. Calculate the offsets per second of the GNSS data between the corrected data and the raw data. Use those offsets to correct the GNSS values in your images in a pseudo differential correction of the camera locations? This is all possible with Python.

Still, the GNSS Z values on your phone camera are going to be values above or below the WGS84 ellipsoid. This might not be useful to you but still, you could extract out the locations of the images and use the XY values, and some existing surface models data to get a better Z values. Again, there is a Python solution for this.

Of course, all of this assumes you are doing this in some location with existing surface model data of sufficient quality, and access to a base station to correct you GNSS data. And finally, what makes you think Google Earth data are 'accurate?'

Improve this answer
1
  • Thank you. My task is basically to use a camera in a city and then find out where that picture has been taken in relation to a map using computer vision. If i would use Google Earth as a reference map, using Google Earth data would be accurate enough!
    – nona
    Commented Feb 7 at 10:39

Not the answer you're looking for? Browse other questions tagged or ask your own question.