Hi all! We are running very large projects with over 3000+ photos shot between a Mavic 2 Pro and a Nikon d5600 with a topeka wide angle lens
However we are having a lot of major offset issues with our manual control points, we beleive it is because the Nikon is in the sonsor database and the mavic 2 pro is not.
Does anyone have the correct data to add the Mavic 2 pro to the Reality Capture Sensor Database ?
Its the most popular pro drone by far and I am sure having the correct sensor data can only help us?
One problem with using the Mavic 2 Pro is that it has a rolling shutter. A rolling shutter can cause issues of distorting and is not well suited for photogrammetry purposes. The rolling shutter effect becomes significant when flying the drone fast and or when flying at low altitudes.
I would recommend using the DJI Phantom 4 Pro because it has a mechanical shutter, and it is back on the market because of the demand for a drone with a mechanical shutter.
i understand the differences but the Mavic 2 pro has a better camera and takes better images - other photogrammetry apps support it directly in their database to address the rolling shutter issues since its the top drone on the market.
are there any plans to add it to the reality capture sensor database ? it would help with processing images and alignment issues etc for sure and i am sure its not hard to do if you know how to do it…
I would really like to understand what CR devs advise when it comes to the DJI Mavic 2 Pro’s built in lens profile? If you shoot JPG, it will be baked in. If you shoot RAW (you’re shooting RAW right?) you have an embedded profile which Adobe products will bake into the image you output for processing in RC. For instance, DNG->Lightroom->tiff->RC. The embedded profile corrects distortion and crops the image, doesn’t this introduce error? So if DJI Mavic 2 Pro is to be supported by RC, what image workflow is supported?
I have been stripping the profile before going into Lightroom, then doing color calibration in Lightroom, and exporting 8bit images for RC to process. I typically find about 3-4% reduction in error compared to profile being baked in.
Dear ,
The inputs can be in any of the supported formats, but I recommend you to shoot RAW decrease highlights, raise shadows, adjust the exposure and export JPGs into RC without doing the lens correction outside of RC, it is best to let RC do it when it has the profile information.
Thanks for replying … I am well aware of the issues doing lens correction outside of RC and I thought I made that clear in my post, so that makes me wonder if you understood my above post. Let me try again… The images from the Mavic 2 Pro already have lens correction on them!
If you shoot JPGs with the Mavic 2 Pro and bring them into RC, the image will be undistorted (lens correction) and cropped.
If you shoot DNGs (RAW) with the Mavic 2 Pro and load these into Lightroom or Photoshop through CameraRaw you will get the the same undistorted (lens correction) and cropped image.
Both scenarios are NOT best for RC! This is why I have been stripping the metadata of the DNG files to remove the embedded lens correction profile. Then doing processing in Lightroom and exporting to RC. I have found about a 3-4% reduction in median and mean error rates. This may not be a lot but it is consistent reduction of error for me when I remove the embedded lens correction.
What do the developers think about this? You can’t just put in a ccdWidth of 13.2 mm in the sensor database if the image is being heavily manipulated by default. Even if you guys are suggesting to shoot RAW and do only color processing you STILL get a lens corrected image out of Adobe products.
Oh okay , thank you for clarifying this to me. That is very odd but I will surely communicate this with my colleagues, Thanks again for providing us with such information.
Please let me know if there is any information I can provide to help with this. Or maybe I don’t understand how RC internals work that make this a non-issue? I am hoping at the very least there is just an official word from RC Devs on how to handle Mavic 2 Pro data (or any DNG with an embedded profile) to minimize error.
Hello , Does the development team need any more info or data on this subject? Here is an example of the crop and undistort that happens with the embedded profile.
What was the outcome of this? I’m shooting JPEG because the automated mapping apps will only allow this. Do I need to paste <camera model=“DJI L1D-20c” ccdWidth=“13.2”/> into the database or will this cause errors because the JPEGs from the drone already have been corrected?
Hello Chris,
If your images are already corrected automatically then do not. If you have these images with the distortion and have not been corrected yet, you can try the method. There is an article in the app HELP called Extending sensor database I believe, where it is further explained on how to add the sensor information manually.
It seems that meanwhile it is not longer possible to manually add a new camera to the database. There is no database anymore in the program files - at least not in the PPI version.
