Which picture ratio?

[andy_k]

I wouldn't expect you to have any issues shooting landscapes or shots with strong straight lines especially shooting straight down - I excluded those from the 'problematic' class earlier

The distortion becomes far more apparent when you shoot complex urban environments (dense housing) at an angle of around 30-45 degrees.

 

[With The Birds]

it's not the 'extreme edges - it's about 20% of the image - check my post where I showed the two different output



Not quite so simple, the correction algorithm isn't just a crop, it's scaling/warping parts of the image (again see my previous posts) so when it comes to stitching, even the middle areas of the frame have some distortiion. On expansive landscapes it doesn't matter, on shots where there are lots of string verticals, it's easily sorted but in complicated urban landscapes a lot of manual correction is needed if you don't want duplicate houses etc in the final panorama



OK, let me try and explain what I'm saying

As (some of us seem to think) the 16:9 utilises more of the original raw data (not the DNG) and that it sometimes renders a wider view in the same final format it effectively has a (albeit slightly) wider field of view.

Close to the camera that could result in images just a couple of centimeters wider but as you move further away from the camera that gap between the two get's much wider - at 3m a couple of centimeters could easily become dozens of meters as your target moves further away from the camera.

Because the camera is correcting images prior to saving them it appears that it's not 100% accurate and we do get some variation in the perceived focal length

Ok- so you are saying the issue really is that the correction applied in camera with the intent of compensating for light fall off and distortion isn't consistent. To the extent that is the case then yes stitching will be more of a challenge and the amount of error might provide that an acceptable result can't be easily achieved with some image sets (presumably as the masking/blending warping of pixels will leave obvious artifacts).

This probably shouldn't be a surprise, we have what is essentially a CCTV camera standard lens in the P4P. At the price/performance point there will inevitably be compromises. Those who need consistent professional grade results with less effort in post and shooting have options in the inspire 2 and other AC.

Your explanation re FOV and apparent size of objects in frame is well understood, doubling or halving of the lens to subject distance will halve or double the apparent size (area occupied by object). It is likely irrelavent to the discussion. The fact probably is that the in camera corrections applied are consistent and that less complex scenes are less subject to obvious stitching errors.

 

[SteelFlyer]

Stitching issues?

Perhaps you should try another stitching program?

What stitching program are you using?

 

[Meta4]

What stitching program are you using?

Autopano Giga and Microsoft ICE

 

[AyeYo]

Stitching issues?
About half of my shooting is panoramas and I haven't encountered these stitching issues.
This example is stitched from 10 individual images.
Perhaps you should try another stitching program?

I'm referring to 3D/VR panoramas and I think others are as well.

 

[With The Birds]

I'm referring to 3D/VR panoramas and I think others are as well.

Probably a very useful clarification from you here. I wouldn't have picked that you, or others from that matter, had VR/3D in mind as the usage case. From your opening post you wanted to learn which image format provided the "fullest" or largest image- 3:2 is the obvious answer (the only presentation that might use the full sensor). VR/3D require specialties camera rigs to do properly and we might confidentky say it wasn't a consideration when DJI specified the imaging system for the P4P. Unfortunately your only option to peruse this field is a larger AC flying a 3D/VR imaging system. You could probably get a convincing effect shootimg stills moving the AC slightly between frames with the right post software.

Please tell us more about your usage case and workflow.

 

[andy_k]

The fact probably is that the in camera corrections applied are consistent and that less complex scenes are less subject to obvious stitching errors.

The problem I am seeing is that they don't appear to be consistent, the 'real' output seems to vary between the equivalent of 15mp and 17mp - when this is subsequently scaled up to the claimed 20mp it means some frames are being scaled by 25% and some by less than 20% - this discrepancy means (in an overlapped panorama series) that the 'overlap' can be two completely different scales.

It's a very odd set up and my request for a detailed explanation of the process from DJI is still unanswered.

 

[andy_k]

You could probably get a convincing effect shootimg stills moving the AC slightly between frames with the right post software.

Please tell us more about your usage case and workflow.

You should try that.
You could also try the same by using individual frames from a movie.

You may find that the results don't match your expectations

 

[With The Birds]

You should try that.
You could also try the same by using individual frames from a movie.

You may find that the results don't match your expectations

Luckily I won't be disappointed as I have no interest in producing 3D images.

 

[ryantrax]

Also using 16:9 images can mess up the stitching of 360 VR Panos like this:
Downtown Dayton, Ohio | Virtual tour generated by Panotour

 

[DefiantChild]

Also using 16:9 images can mess up the stitching of 360 VR Panos like this:
Downtown Dayton, Ohio | Virtual tour generated by Panotour

Shameless plug aside, where are the stitching issues in the linked 360?

 

[ryantrax]

Shameless plug aside, where are the stitching issues in the linked 360?

LOL, no plug, just clarifying what I meant by the 360 VR panorama. There aren't any issues on the one I included. The software has a hard time stitching when you use 16:9 images because it crops the top and the bottom of the image, then there isn't enough overlap.

 

[Shammyh]

So, for those of us wanting to use the P4P for photogrammetry... What's the verdict on all this?

I took a look myself, and indeed, it looks like Lightroom is cropping/warping quite a bit of the actual RAW file.
Doesn't seem to matter terribly for landscape photos (or just regular photography in general) but for those of relying on accurate sensor readings... this seems problematic.

~~~

So as to how to deal with this moving forwards:

Should I set up my own consistent (and perhaps less severe) lens correction in RawTherapee and use that to export PNGs/TIFFs for Photoscan/etc use?

Should I rely on the DJI-created Lightroom-style cropping and use that? Or should I just let it produce it's own JPGs and use that?

For a photoscan/DroneDeploy/Pix4d style downward-facing photogrammetry workflow, what's best?

Just curious if anyone has any thoughts on these questions... Thanks in advance!

 

[Meta4]

So, for those of us wanting to use the P4P for photogrammetry... What's the verdict on all this?

See post #80 above
If you use DroneDeploy it will shoot 4:3

 

[AyeYo]

Probably a very useful clarification from you here. I wouldn't have picked that you, or others from that matter, had VR/3D in mind as the usage case. From your opening post you wanted to learn which image format provided the "fullest" or largest image- 3:2 is the obvious answer (the only presentation that might use the full sensor). VR/3D require specialties camera rigs to do properly and we might confidentky say it wasn't a consideration when DJI specified the imaging system for the P4P. Unfortunately your only option to peruse this field is a larger AC flying a 3D/VR imaging system. You could probably get a convincing effect shootimg stills moving the AC slightly between frames with the right post software.

Please tell us more about your usage case and workflow.

This is completely false. I can shoot VR panos with perfect success using everything from a crappy Mavic camera to my cell phone. The P4P specifically causes issues because of the crop and rescale process distorting the photo.

 

[AyeYo]

So, for those of us wanting to use the P4P for photogrammetry... What's the verdict on all this?

Wait for Yuneec H520 release. Buy H520. Profit. The P4P will be my photography drone and not much else.

 

[AyeYo]

For a photoscan/DroneDeploy/Pix4d style downward-facing photogrammetry workflow, what's best?

This I actually haven't had an issue with. Whether Drone Deploy and Altizure have better stitching algorithms or whether the P4P's distortion doesn't affect flat image stitching as badly as equirectangular, I'm not sure. My 2D and 3D maps with both Drone Deploy and Altizure come out great. Whether they're survey level accurate I don't know and don't really care, as that's not the type of thing they're being sold for. So, no, you shouldn't have any major issues with the P4P in those areas.

 

[With The Birds]

This is completely false. I can shoot VR panos with perfect success using everything from a crappy Mavic camera to my cell phone. The P4P specifically causes issues because of the crop and rescale process distorting the photo.

It seems the P4P images aren't the only thing being distorted- you won't shoot 3D images with any of your sighted hardware with any success. Read your post I responded to, you clearly stated "3D/VR".

 

[AyeYo]

It seems the P4P images aren't the only thing being distorted- you won't shoot 3D images with any of your sighted hardware with any success. Read your post I responded to, you clearly stated "3D/VR".

That's odd, because I've been doing it for months. 3D pano, aka 360 pano, aka VR pano. Not a stereoscopic pano.

 

[With The Birds]

That's odd, because I've been doing it for months. 3D pano, aka 360 pano, aka VR pano. Not a stereoscopic pano.

Yes, it is possible. I can produce a colour image from three seperate monochrome images taken with colour filters also. You might get some very pleasing results however you won't approach the consistency of output and convenience arrived at by using dedicated imaging hardware.