Forum

Author Topic: Ultra wide lens choice  (Read 7862 times)

ThomasVD

  • Full Member
  • ***
  • Posts: 107
    • View Profile
Ultra wide lens choice
« on: November 16, 2016, 07:24:33 PM »
Hi everyone,

I'm looking to buy an ultra wide lens for my Nikon D5300 (DX) camera, to use exclusively for photogrammetry (especially in scenarios with limited space). I understand that PhotoScan's algorithms assume a 50mm equivalent and Brown's calibration model, but I guess that as long as the lens doesn't create too much image distortion the software should have no problem aligning and reconstructing a scene?

I spent a lot of time Googling and have narrowed down my choices to these 4 options:
- Tokina 11-16mm - this one seems to be a customer favourite on online forums but gets slightly higher distortion
- Tokina 11-20mm - this one gets the best DxO rating with my camera
- Tokina 16-28mm - according to DxO this one has the least distortion
- Sigma 10-20mm - widest angle, gets good reviews, but apparently less sharp along edges

Each gets great reviews, but of course none of those reviews take into account photogrammetry!

Which would you recommend? I'm looking for something as wide as possible which still gets good photogrammetry results. I've taken into consideration sharpness, auto-focus enabled, price, amount of distortion - am I missing anything? Is there some reason any of these would simply be a bad choice for photogrammetry?

Any feedback is as always much appreciated!

Cheers,

Tom
« Last Edit: November 16, 2016, 07:31:55 PM by ThomasVD »

James

  • Hero Member
  • *****
  • Posts: 736
    • View Profile
Re: Ultra wide lens choice
« Reply #1 on: November 17, 2016, 03:04:43 PM »
I think photoscan only assumes 50mm in the absence of any exif data suggesting what the actual focal length is.

Lots of distortion isn't necessarily a problem as the distortion model will compensate for it, but 'weird' distortion might be a problem. i don't know how you quantify that without running the photos through photoscan and looking at the calibration report though!

I would look at fisheye lenses for very tight spaces. Nikon or Sigma have done well for me in the past, and you are probably better off with full frame (as opposed to circular - nothing to do with sensor size) as it uses all your pixels although you don't get quite such a wide field of view as with a circular fisheye. also with a circular fisheye you need to mask out the big black round 'border' in pscan which is a pain anyway.

some example stuff here: http://www.agisoft.com/forum/index.php?topic=3307.0

David Cockey

  • Full Member
  • ***
  • Posts: 207
    • View Profile
Re: Ultra wide lens choice
« Reply #2 on: November 17, 2016, 06:14:55 PM »
I've gotten excellent results with Photoscan using a Canon EF-S 10-22 mm zoom lens on a "crop" sensor Canon camera (16-35 mm full frame equivalent). On a couple of boat projects there was very limited space around and under the boats and capturing the boats would not have been feasible without the very wide angle. Photoscan determined the lens distortion coefficient and corrected the lens distortion in the same manner it does for any lens.

Obviously the Canon lens isn't an option to use with on your Nikon camera. Before purchasing the Canon lens I considered the the other lens mentioned above. I would go with one with 11 mm minimum focal length rather than 16 mm. The difference in angle of view is huge. Also my recollection is one has more "complex" distortion which might be more difficult for Photoscan to completely correct.

ThomasVD

  • Full Member
  • ***
  • Posts: 107
    • View Profile
Re: Ultra wide lens choice
« Reply #3 on: November 17, 2016, 06:39:04 PM »
Thanks James and David,

I can always rely on you for great to-the-point advice, you're a big part of why these forums are still so helpful!

I considered fisheye, and I have indeed seen some of the excellent work posted by users such as BigBen, but I'm inclined to buy a normal ultra wide lens because it probably plays nicer with other photogrammetry software I sometimes use (ie not every software is so good at working with fisheye as PhotoScan is).

I'm also inclined to go with as broad a field of view as possible; the reason I included the 16-28mm is because I know Wishgranter has had good experiences building photogrammetry models with this lens, whereas the other 3 are sort of a shot in the dark where photogrammetry is concerned.

David you mention that one of them may have more "complex" distortion - I remember reading a recommendation by Alexey that the lens distortion should be compatible with Brown's model, however I have no idea how to determine whether or not a lens conforms to Brown's distortion model. However the DxO Mark pages I link to do show quite detailed distortion information (profiles / grid at various F-stops) - is there any way this can help in determining whether or not a lens is suitable for photogrammetry?

Kind regards,

Thomas

Kiesel

  • Sr. Member
  • ****
  • Posts: 297
    • View Profile
Re: Ultra wide lens choice
« Reply #4 on: November 19, 2016, 07:53:35 PM »
Hello Thomas,

some words also from my side.

Generally, as you know for sure, all zoom lenses are a compromise and especially for photogrammetry. When ever possible and for best results it is recommend to use a prime lense.

I have only found one ultra wide lens from Nikon:
- Nikon AF Nikkor 2,8/16
one from Sigma:
- Sigma 14mm F2.8 EX Aspherical HSM
and three from Samyang (not sure, but I think all of them have manual focus):
- Samyang 10mm F2.8
- Samyang 14mm F2.8
- Samyang 16mm F2.8

Nevertheless to the above ultra-wide zoom lenses have also big potential for photogrammetry and they are mostly even wider than prime lenses!

I use a Canon EF-S 10-22 mm by myself like David Cockey with good results.

So there is an additional one you could choose:
- Tamron SP 10-24MM F/3.5-4.5 Di II LD Aspherical IF
distortion profile on your camera: https://www.dxomark.com/Lenses/Tamron/Tamron-SP-10-24mm-F35-45-Di-II-LD-Aspherical-IF-Nikon-mounted-on-Nikon-D5300---Measurements__919

Very important for photogrammetry is the solidity of the lens. So go to a shop or a rent service and test the lens if you can shake the front element for example and get a feeling of its robustness. Is there any clattern, slackness and so on...? Otherwise you get a mean calibration including all of those instabilities.

One problematic distortion is the so called "moustache distortion" or complex distortion which is a mixture of barrel distortion in the image center and pincushion distortion near the picture margin see https://en.wikipedia.org/wiki/Distortion_(optics). If you have a look at the distortion profiles or distorsion grid at DxO you can see if it is one or not. Normal distorsions can easily approximated by a polynom (Brown model) and the polynom approximation of PhotoScan has four coefficients K1, K2, K3 and K4 for radial and four for tangential distorsion p1,p2,p3 and p4, enough for a very good approximation.

Cheers,

Karsten
« Last Edit: November 20, 2016, 09:16:28 PM by Kiesel »

ThomasVD

  • Full Member
  • ***
  • Posts: 107
    • View Profile
Re: Ultra wide lens choice
« Reply #5 on: November 21, 2016, 02:25:22 PM »
Hi Karsten and thanks for your helpful reply!

There's just too many options to choose from isn't there haha + too many factors to take into account ;)

However like you say the Samyang lenses don't have auto-focus which is a bit annoying (other than that they get great reviews!) and I like that the Sigma 14mm is a prime but it seems to suffer a bit from mustache distortion (see DxO profile + comment nr 1 on http://www.fredmiranda.com/reviews/showproduct.php?product=75).

The Tamron SP 10-24mm has a nice zoom range, but on the other hand it gets a much lower DxO score than the Tokina 11-20mm for my Nikon D5300 (I think it's mainly due to less sharpness + less light transmission). Seeing as the distortion profiles look about the same as the Tokina 11-20mm (even though the Tamron has aspherical lenses which according to Wishgranter are better for photogrammetry) I think I'm going to choose for the latter. According to some reviews the Tokina also has a faster / more accurate auto-focus.

So I guess after about 10 days of forum / Facebook / Google searching that's my mind made up, let's hope it's not a horrible decision haha. If I remember I'll post some results here once I've received the lens!

Thanks everyone for the input and feedback!

Cheers,

Thomas

James

  • Hero Member
  • *****
  • Posts: 736
    • View Profile
Re: Ultra wide lens choice
« Reply #6 on: November 23, 2016, 09:28:43 PM »
Not much to add, and probably not helpful if you already went for tokina, but I do own the tamron 10-24 (probably am older version) and have happily used it with photoscan on my Nikon d90.

Can't say much about the sharpness, as I'm normally just happy if my images align, but they did well :)

I think this may be the only thing I actually shot with it for scanning:
https://skfb.ly/CosH

DivingArchaeology

  • Newbie
  • *
  • Posts: 1
    • View Profile
Re: Ultra wide lens choice
« Reply #7 on: November 23, 2016, 10:14:09 PM »
Hi Thomas - I know you made your purchase, but thought I would put in my two cents:
I have a Tokina 11-16 DX II mounted on a d5300 which I use for both terrestrial and underwater photogrammetry. I have been fairly pleased. The distortion is no problem for photoscan - I get excellent, highly accurate results. A very high degree of overlap is key. The zoom is a mixed bag. I know there is less chance of accidently zooming if you set it at 11 or 16, but usually pull back to 13-14 mm and hope I don't bump it - that range works best based on my EQ setup, distance from the ground, and coverage pattern. My main complaint is that I wish the lens was sharper. I am tempted to replace it with a 11-20 (or a prime) - so congrats on your purchase!

ThomasVD

  • Full Member
  • ***
  • Posts: 107
    • View Profile
Re: Ultra wide lens choice
« Reply #8 on: November 30, 2016, 05:19:37 PM »
Thanks for the additional info guys! As promised a little review of my experience with the Tokina 11-20mm so far! (I'm working with a Nikon DX camera, so this is roughly 16-30mm range for a full frame camera).

A first pleasant surprise: the box in which the lens came states that it's an aspherical lens! Hadn't read that in any of the reviews or online specs, but aspherical lenses are supposed to be perfect for photogrammetry so that's great.

Secondly: ultra-wide really does mean ultra f*cking wide. I guess I theoretically "knew" that the angle was going to be very wide, but I didn't really grasp just how wide until I took my first pictures. You can literally photograph another person standing at arms length and have their entire body in that single shot, with plenty of ground and sky left along the edges. My first tests were on a square with a church, and I could easily capture the entire opposite side of the square with a single picture, whereas with my regular 35mm 'wide' angle lens I'd need at least 8 pictures (4 along the width of the square + a second row of pictures above that to get all of the facades).

In terms of handling: the lens feels very sturdy, the zoom ring is quite "resistant to change", so you don't have to worry about it moving in our out without you actually meaning to do so, and the auto focus is fast and accurate.

Picture quality looks great (here's a landscape shot I did, edited in DxO), the wide aperture allows for good shooting in low light and at F11 the images look sharp from corner to corner.

So my first photogrammetry tests were on a city square with a church: I first recorded a statue and the bottom of a tree there. I then decided to push my luck and record a tunnel by just aiming the camera in the same direction as the direction of the tunnel (not the walls of the tunnel) and hope that because the view angle is so wide the entire thing would align. Finally I made some shots of the church on the same square. The images were a mix of 11mm, 16mm and 20mm. Total number of images: 173.

For my PhotoScan tests I decided to just align all images together, and wonderfully they aligned all at once, even though I made no attempt to make a good photogrammetry transition between the statue and tree on one side of the square, and the church at the other end of the square. Camera locations shown here. The tunnel is the only part of the scene for which I built the dense cloud in PhotoScan. I was really impressed that the images of the tunnel aligned without problem, considering that none of the surfaces (walls / ground / ceiling) of the tunnel were ever in the middle of the pictures but only at the edges. The dense cloud shows the general shape nicely, and of course the resulting mesh has quite a lot of noise (since all pixels were captured at a very oblique angle, at the edge of the field of view of each picture) but as I said this was just an alignment test so that's fine.

In order to test what kind of quality you can get when the camera is actually angled perpendicular to the surface being recorded I also took 81 pictures of a house. The house is in a relatively narrow street but with the wide angle I had no problem getting the entire building in a single shot from across the street. Again all images aligned without a problem and the resulting dense cloud looks good. If we look at the mesh it's clear the area of interest on which the cameras were oriented is cleanly reconstructed (with the exception of the windows because of glass) - a very large area around the area of interest is also reconstructed but this contains more noise.

So conclusion: I think the Tokina 11-20mm F2.8 AT-X PRO DX lens is a great price / quality purchase for photogrammetrists! The software didn't have any problem aligning the images even when taken at the widest 11mm and the wide angle allows you to significantly reduce the amount of pictures needed to completely record large or cramped spaces.

The only downsides compared to my regular Nikon 35mm prime is that the Tokina is significantly heavier (you definitely can't shoot from just the wrist, you need two hands; that's all right for a couple of pictures but if you're taking thousands of a pictures in a day I can imagine your arm getting tired). You also have to remind yourself that even though an area at the edge of the field of view might be reconstructed, you also have to take images of that surface with the camera oriented perpendicular to the surface (nadir) in order to avoid surface noise. Finally the Tokina captures a lot of the scene in a single shot, but those pixels still have to share the same sensor space (in my case 24MP), so you can't expect the same mesh detail from 20 ultra-wide pictures as you'd be able to get with 160 less-wide closer up images of the same scene taken with another lens.

I hope this little "review" is helpful to someone - might write a more scientific review once I have some more time!