What is the feasibility and potential risks of using ODM to process aerial images captured from multiple drones. The primary concern revolves around camera calibration consistency, especially when using drones with similar or different camera specifications.
Context
In my current project, we have access to multiple DJI Mavic Pro drones. The objective is to utilize these drones concurrently to cover larger survey areas more efficiently. However, we are uncertain about the implications of merging datasets from multiple sources in ODM, particularly in terms of camera calibration.
Questions
Multi-Drone Compatibility: Is ODM capable of effectively processing images captured from more than one drone, specifically when the drones are of the same model (e.g., DJI Mavic Pro)?
Camera Calibration Concerns: What are the potential challenges or risks associated with camera calibration when processing images from multiple drones? How does ODM handle variations in camera calibration data across different drones?
Impact of Using Different Drone Models: How would the scenario change if the images were captured from different drone models? Would this pose additional challenges in terms of camera calibration and image processing in ODM?
Best Practices and Recommendations: Are there any best practices or recommendations for undertaking such a multi-drone survey approach to ensure optimal data quality and processing efficiency?
I have used a Mavic 2 Pro and Mavic 2 Zoom together to map out property, and didnt notice any issues in the ortho generation when mixing all of the 2000+ images together.
Sam_G has provided a practical perspective which is a good foil for my theoretical perspective:
With those particular drones, given they don’t have fixed focus, the within camera variance in calibration coefficients is likely as large or larger than the between camera variance, and so it doesn’t matter or doesn’t matter much if they might be treated as the same camera.
Sam: in your report, do you see multiple camera calibrations or one for all cameras of the same model? If the former, there is nothing to do. If the latter, then @lupion_pe, you could consider modifying exif tags to ensure that each camera is separately calibrated.
As to multiple camera models, ODM handles that quite well. In principle, your models are more likely to be more accurate with multiple camera models, as your likelihood of bias decreases, i.e. discerning 3D scene characteristics from camera calibration becomes easier / more robust. Or to put it another way, balancing intrinsics and extrinsics is more robust with multiple camera models.
As a matter of fact, I do see 2 different camera model details. I had not even noticed that until you asked.
Hasselblad for the M2Pro and DJI for the Zoom.
If it would be of any use, I can provide screen shots of them
I’ve used images from 2 M2P drones in a single reconstruction with no problems. Whether or not there were multiple camera calibrations I can’t remember.
I’ve also combined drone and DSLR images with no problems.
