Have gone through all I could find on calibration of drone cameras. I have a question now, or several.
Each time I run a new set of photos then look at the camera.json calibration file, the numbers are different.
How much processing does it take to compute this? That leads to the use of a standard camera.json file loaded into the parameters. Would it give more consistent results or lower, enough to matter, the processing time required?
If I were to take a number of camera.json files, put them all in Excel, average all the rows, then use that would it give me better results? Or does the photo capture play into the calculations for each set, so it is not always a “once and done” calculation?
Thanks, G
2 Likes
I don’t think the camera model is terribly computationally intensive, so I don’t think you’re saving much time there.
It will vary each time as literally nearly everything impacts a camera calibration model, including air temperature at the time of collection.
You could, in theory, derive a pre-set model, but I don’t think it is going to save you time. You’d mostly want one if you’re having particular difficulty getting a good calibration over a super-flat field or something.
1 Like
I find that the residuals grid, which shows distortions across the FOV, varies quite a lot between tasks, so it’s probably not a good idea to use a previously determined or average model.
It will also vary depending on environmental factors as mentioned above, and also it’s going to be dependent on method of feature extraction. ORB for example concentrates on the middle of the FOV, mostly ignoring features towards the edge.
1 Like
Thanks for the responses. Seems like leaving it alone is the best course of action. Was just exploring the different settings available and saw the option to use a saved file.
Thanks again!
2 Likes
Keep it in your toolkit for sure 
Hopefully we’ve gotten to the point that you’ll never need to worry about passing a manually-calibrated lens profile in normal usage, but there are certainly edge-cases where it helps.
1 Like
If you have a fixed camera, with global shutter, no focus, no zoom, and it’s designed to keep the same characteristics across a range of different temperatures, then you can calibrate once and use that for a long time.
Unfortunately, the list of cameras that qualify for this is quite short, and generally run very in the 10s of thousands of dollars. And they’ll likely come with a set of calibration parameters. The rest of us probably need to self calibrate each time.
2 Likes
Hmm. Most of my ODM runs contain several individual missions, all taken within the same 1-2 hour window. Will the differences average out when the camera calibration is run against photos from all three flights?
1 Like
It could, and it could not. It is something I’ve been meaning to experiment with, but I’ve not found the time to spend that much time in the field with it.
1 Like
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.