Modelling an engine block

[Jokeri]

Hello!

I dont seem know how to use the program properly.
As the title says I'm trying to create 3D model of an engine block.
I've tryed to build the model with and with out masked background and it really doesnt make the difference. (Could it be that the engine block is too shiny?)

If I could create model like Erik Leeman has it would be great
http://www.flickr.com/photos/erik-nl/7084218817/
Unfortunately I lost my hard drive, so I dont have any pictures of the agisoft program, but the created surface is like watching the mount everest from above..

Can someone help me out, you can get the pictures from the links below..

Two cameras has been used Canon 600d and Canon 5D

https://drive.google.com/file/d/0ByPwUKOuT_HBUWZMLVRCaFJWY1k/edit?usp=sharing
https://drive.google.com/file/d/0ByPwUKOuT_HBbFVMNzdQNEhtcVE/edit?usp=sharing
https://drive.google.com/file/d/0ByPwUKOuT_HBZjlzZC1MSGRWWTg/edit?usp=sharing
https://drive.google.com/file/d/0ByPwUKOuT_HBZnY0Qzd1V3FvMTQ/edit?usp=sharing

[Matt]

Shiny/reflective is bad. You will probably need to coat it with something dull to get the photo matching to work.

[James]

Never under estimate photoscan's magical abilities to align images and extract depth data where the human eye can not perceive it!

I didn't bother trying with the f1.8 images but the f16 seem decent. maybe somewhere in between like f11 you could get a quick enough shutter speed to completely eliminate camera shake which was apparent in some of your images at 1/40-1/50s a still get enough dof.

the image below shows my results and all settings which i think are better than mount everest although still need a lot of tidying up.

first i calculated image quality and disabled all images below 0.55 as these showed a lot of camera shake blurring. i then aligned on high (dont bother masking) and then removed double images (removing the image with lowest image quality - check the image list in my screen shot). I cropped the sparse point cloud down to the engine block area.

then calculated dense point cloud on high, and again cropped down to the engine block and removed any crazy bits.

calculating mesh on high used all of my 16gb ram.

the insides of the cylinders are not reconstructed at all, but the exterior shape is pretty good.

[Matt]

You did well with that  .  What blur algorithm are you using?

[James]

it wasn't much to do with me, i didnt think it would work to be honest!

photoscan 1.0 has an estimate image quality function (right-click on images in the photos pane) that estimates image quality based on sharpness so you can easily spot 'not so sharp' images, if that's what you mean?

using build 1745 by the way, not tried 1760 yet.

[Jokeri]

Thank you for your help, I got it now.
All I'm missing is the full version of the program

[Kiesel]

Hello Jokeri,

do you mean with

All I'm missing is the full version of the program

the professional version of Agisoft PhotoScan?

You don't need it for your engine, you can also do it with the standard version and scale your results for example with the free Meshlab afterwards .

As Matt suggested

You will probably need to coat it with something dull to get the photo matching to work.

use for example some Anti Flare / Dullingspray. You can also improve the photos/results further by projecting some noise pattern on your dull object.

Karsten

[EMULAT3D]

While we are talking about capturing reflective surfaces with PS, have you guys seen what Ten24 has been able to do?

Non Destructive Reflective Surface Scanning using Photogrammetry
https://www.facebook.com/media/set/?set=a.10151518080817263.1073741840.192982602262&type=1

Apparently this isn't achieved by using polarization. HDR's maybe to reduce blowout? Anyone have any ideas?

[Wishgranter]

we have discussed it with AGI team. it seems that the meshes are IMPORTED back from something other than AGI..... post-work possible. But TEN24 could say something to that :-D

[chadfx]

Looking at their source photos here:

http://www.facebook.com/photo.php?fbid=10151519321267263&set=a.10151518080817263.1073741840.192982602262&type=3&theater

It doesn't seem like they are doing anything special (highlights still blown out, deep shadows, no polarization, etc). There are some pretty big gaps in their point clouds and very loose geometry in those highlight areas as you would expect. So they certainly might be doing some of the work outside of AGI.

That being said, I have been pleasantly surprised on a number of occasions how well AGI can deal with highlights that are moving around in the photo survey. It even has filtered out reflections from glass museum cases. (faint reflections, mind you...but still nice to see).

[Alexey Pasumansky]

Non Destructive Reflective Surface Scanning using Photogrammetry

Hopefully, Ten24 will reveal the mystery one day =)

For small objects I think fixed camera rigs could be used: one shot of the object with projected texture, covered with flour or spray, and the second shot - of the clean object for texturing purposes. Thus swapping image folders it would be possible to get good textured model, of course, providing that the object has not moved during the mentioned operations.
But for big objects like cars this method will require huge resources, compared to full head/body scanning, as I assume.

[CjC]

Do you guys think ten24 used a hand scanner and it was imported from artec studio9?

[FoodMan]

I've seen a screen shot of ten24 car scan, and you can see the photos all show heavy reflections, so clearly these photos are not the ones they used to reconstruct... and I don't thing they used the artec either, because re-importing the artec scan into Pscan would be a nightmare... and pointless if the goal is just texturing..

Going to search a bit.... hehe

f/