Messages

1

General / Re: Shooting interior with Nikon D5300

« on: October 31, 2022, 06:49:18 PM »

the lens is AF-S DX Nikkor 18-55mm f/3.5-5.6G VR II. I really why the images are blurry as I am using a tripod with remote shooting using my smart phone connected by Wifi to the Nikon... quite baffling...

like you say, something doesn't seem right.  is the lens and the sensor clean?

maybe try with the Ricoh and reassess

2

General / Re: Shooting interior with Nikon D5300

« on: October 31, 2022, 05:59:32 PM »

The Nikon D5300 I was testing has a Zoom 18-55 mm lens set to 18 mm....

this one?

https://www.nikon.co.uk/en_GB/product/nikkor-lenses/auto-focus-lenses/dx/af-p-dx-nikkor-18-55mm-f-3-5-5-6g-vr

i don't think it should be anywhere near as bad as your results show.  maybe there's something wrong with your technique?

when you took the interior shots are you very careful pressing the shutter button (so as not to move the camera)?

have you tried using a shutter delay to take the shots? (e.g. 2 second delay)

i would make sure there is nothing wrong with your current technique before buying more equipment

3

General / Re: Shooting interior with Nikon D5300

« on: October 31, 2022, 01:59:12 PM »

I am coming to the conclusion that the Nikon I used has some heavy blur problem, given that I was shooting at f8 with tripod and remote trigger

which lens are you using?

I will look at the Ricoh GR for my  next tests....

Ricoh GR is a nice camera, but it doesn't have IBIS (in body stabilisation), so you will need a tripod with interiors

If you could tell me a little bit more the camera, settings and interior lighting conditions

lighting conditions - available light only.  turned on the lights in any darker rooms, otherwise just daylight through windows

camera - i used a Pen F, but any MFT camera with IBIS and interchangeable lenses will be fine.  if you just want the cheapest body that will do the job, then take a look at a used EM5 mk1.  that was the first Olympus camera to introduce the modern 5 axis rated @ 5 stops, which is the same as i have.  Used ones can be found for <£150 (e.g. https://www.mpb.com/en-uk/product/olympus-om-d-e-m5)

lens - Samyang (also known as Rokinon) 7.5mm fisheye (FE).  manual focus, but that's not a disadvantage.  i tried an AF lens but in practice results were worse and more time consuming because the focus point was often not where i wanted it.  with the Samyang FE you set focus based on the scene, somewhere close to infinity and at an aperture of around f5.6 (equivalent to full frame f11) just about everything is in focus.  the lens is pretty much sharp edge to edge which i guess is why the estimated image quality is so high.  used copies again from <£150 (e.g. https://www.mpb.com/en-uk/product/samyang-7-5mm-f-3-5-umc-fisheye-micro-four-thirds-fit)

don't take my word for it though.  this is just what i use and i get great results for my use.  pls do your own research though

4

General / Re: Shooting interior with Nikon D5300

« on: October 30, 2022, 01:22:55 PM »

JJ, our scans  are for eventually to generate precise point clouds of the interior of a Paper Plant. I agree with you that white even colored walls and ceiling are really not good for photogrammetry...

my 2c

sharp photos depend more on the quality of the lens than the quality of the sensor.  i mentioned this is another thread but don't think anyone took it seriously, because most FF and/or APSC users dismiss smaller sensors without taking the time to understand the benefits

i use a camera pole not a tripod, even for interiors, and the only reason why i even use a pole is because it makes it easier to reach over things - reach over plants and low trees, reach over tables, sofas etc.  for exteriors the pole also makes it easy to take photos higher up (i scan everything from the ground)

i normally work with exteriors but a recently started a redesign job of a house, both interior and exterior work.  i don't need detailed scans because i rebuild everything in CAD (i use the scan to create reference floor plans and to confirm dimensions and levels), but i think my observations are useful because by simply increasing the number of photos, better results can be had

this was my first interior scan, and the house it quite large and has 3 floors, including a bedroom in the loft.  i scanned the whole house in one go - just over 1000 photos in total

tbh i didn't think it would work, so i didn't put that much effort into it, and if i did another one i think the results could be improved significantly.  the only place where Metashape struggles (as i mentioned before) is which plain white walls and ceilings, so i would need to find some way to improve that.  if your interior has plenty of detail on the walls and ceilings i don't think you will have problems

regarding the lens and mft sensor, look at the estimated image quality.  some of the images are shot as low as 1/25 of a second and like i said, no tripod.  in most cases i was holding the pole in my hand off the ground and quickly moving through the house and shooting

i've attached some examples and like i said above, the interior scan could be improved significantly with due care and attention.  one thing i was very happy with is that i scanned the interior and exterior of this house (separately) and later combined them in my modelling software.  the interior model fit pretty much perfectly inside the exterior model

let me know if you'd like to know more

5

General / Re: Shooting interior with Nikon D5300

« on: October 28, 2022, 09:45:49 AM »

what's your use of the scan? a basic model to use as a reference for further modelling or do you need a high quality scan?

one issue will be if you have a white ceiling with little definition

6

General / Re: What is best practice if final mesh has unwanted surroundings next to it ?

« on: October 17, 2022, 09:06:32 PM »

sounds like you use MS similar to me - rough site scan and then rebuild in CAD and/or other freeform modeller?

my understanding is that the SParse Cloud (generated when aligning cameras) will not influence the final mesh or dense cloud at all.  AFAICT they are 2 distinct processes.  this is to align the cameras only so that the next steps can be performed

so the process is ALign Cameras, then optimise them.  how ever few points you're left with will have no effect on the next steps, as long removing points/optimising does not reduce the number of cameras.

once aligned if you generate the mesh from depth maps, the depth maps are created from the photos, not the sparse cloud, so deleting some points will have no effect on the final model.

if you want to remove parts of the model you need to do it after the model has been generated.

the only way to limit the model size before model generation is to rotate/resize the bounding box.

users with more experience will chime in hopefully but this is my understanding

7

General / Re: Best Handheld Camera for Facade photogrammetry

« on: October 16, 2022, 11:31:29 AM »

Andrew,
Something with a large sensor, either full-frame or APS-C (cost-effective/common example: a5100)

i think a micro four thirds (mft) body with a small, high quality, sharp, fast prime lens such as the Lumix 20mm f1.7 (40mm equivalent) would give you clearer, sharper photos than the a5100 and kit lens

in fact you would need the best APSC/FF lenses to compete with the sharpness of the little Lumix

some other advantages of mft:

double the field of view at the same f no (so f4 on mft gives the same DOF as f8 on FF)
IBIS (in body image stabilisation) - best in the industry means you can shoot sharp images at lower shutter speeds
compact and light
high quality small bright prime lenses - cheaper than equally sharp and bright APSC/FF lenses

the main disadvantage is inferior low light capabilities but unless you're planning on doing your facade shoot after sunset, this shouldn't be a problem.  and even then the fact that you can shoot handheld at lower shutter speeds means lower ISO, so lower noise

8

General / Re: Aligned camera vs actual photo

« on: October 16, 2022, 11:12:06 AM »

thanks Alexey - so you mean that the original images may be distorted and that's why they don't quite match?

i use a micro four thirds system which fixes distortion automatically in Lightroom when the RAWs are imported.  there is no way to disable this in LR so the exported jpegs will always be adjusted for distortion - could this affect the results?

9

General / Re: Aligned camera vs actual photo

« on: October 12, 2022, 10:18:23 AM »

anyone from Agisoft care to comment?

does my question not make sense?

10

General / Aligned camera vs actual photo

« on: October 10, 2022, 08:54:36 PM »

trying to develop a workflow where i take a few photos in succession, including my 'hero' shots, then align in MS and generate a rough model which will only act as reference, and possibly as hidden geometry that will cast shadows on the the photo backplate

i took 22 shots of this shop facade.  in MS i aligned the photos (high setting) and then optimised.  once optimised i generated a model from depth maps (high setting, high poly)

i then exported an FBX (with cameras) and opened in MODO.  in MODO i view the scene through one of the cameras and compare to the original photo.  as you can see the result is good but not perfect (if you flick between the 2 images you will see that the alignment is good, but the position is slightly off

the sensor sized was matched in MODO, as was the focal length.  render resolution was matched to the size of the original photo

i've only tried one camera view so far, but will check a couple of others

how accurate can i expect alignment to be?

are there any other settings i should check?

in general will generated models viewed through aligned cameras match original photos?

11

General / Re: different Fisheye distortion effect on alignment and model generation

« on: October 05, 2022, 01:20:12 PM »

thanks Kiesel

I think that's not a good idea to work with autofocus instead of fixed focus

when the lens arrived and i was comparing the 2 i realised exactly this so returned the AF lens

in any case the little MF Samyang i have is sharper and as a bonus it looks cooler too

I would set the focus to something like your middle distance (hyperfocal value) and tape this distance down and not to infinity b.t.w..

agree again.  when i said "to infinity" i actually meant setting the focus to something suited to the site i was capturing

12

General / different Fisheye distortion effect on alignment and model generation

« on: September 28, 2022, 10:49:14 AM »

i use a FE to scan house gardens from ground level and for my use am happy with the results

right now i use a Samyang 7.5mm manual focus lens on a micro four thirds body.  i just set the lens to infinity and pretty much all images are sharp but i avoid moving in closer to the subject

i'd like to change to an auto focus lens to have more flexibility and was looking at the Panasonic 8mm

looking at this review which compares the 2 lenses, you can see that the distortion is different, so to my question - will this difference in distortion be a problem (i.e. will it have a detrimental effect on alignment and model generation) and am i better off sticking with my current lens which is tried and tested?

see here some comparison images:

https://m43photo.blogspot.com/2012/02/samyang-75mm-f35-fisheye-lens-review.html

also:

https://4.bp.blogspot.com/-di4jnTK_rNQ/ULY80ZlCMII/AAAAAAAADK4/fPNhidGu2n4/s1600/fisheye%2Bsharpness%2Bsamyang%2B7.5mm%2Bf3.5.jpg

vs

https://1.bp.blogspot.com/-jhAGnZudzoI/ULY89pHuLxI/AAAAAAAADLE/yLNcaWzuKUM/s1600/fisheye%2Bsharpness%2Blumix%2B8mm%2Bf3.5.jpg

13

General / Re: Build texture for edited and re-imported mesh

« on: June 30, 2021, 10:17:27 AM »

i haven't used Blender specifically for this but i can't see it being any different to other apps.  you just need to make sure you don't move, scale or rotate the model in any way in Blender.

also make sure the model isn't rotated or scaled on import into Blender, or when exporting.

use the 1,3, and 7 (5 to toggle ortho/perspective) keys in MS to see the model in front, side and top view respectively, and you should see the same model rotation in the respective Blender views on import

14

General / 5950x heat?

« on: June 20, 2021, 11:02:13 PM »

those of you that have this CPU, what are your temps like at idle and full load?

15

General / Re: build mesh from depth maps vs dense cloud

« on: June 20, 2021, 11:00:16 PM »

thanks for the suggestion James.  i went through the process but using polygon size as a filter doesn't work as the parts i need to delete don't necessarily have larger polys.

so AFAICT it's either mesh from Depth Maps which is better defined mesh but unwanted extrapolated polys or mesh from dense cloud which will give a better cropped mesh, though not as defined.