Results of combining 3D-scans and photogrammetry
- arie
- I have made 40-50 posts
- Posts: 47
- Joined: Wed Jun 05, 2013 11:43 am
- 10
- Full Name: Arie Kai-Browne
- Company Details: University of Applied Sciences Berlin
- Company Position Title: Scientific staff member
- Country: Germany
- Skype Name: aaaarie
- Linkedin Profile: No
- Location: Berlin
- Been thanked: 11 times
Results of combining 3D-scans and photogrammetry
Hello everybody,
here are some results of a small project I did within my seminar "3D-documentation" at the University of Applied Sciences, Berlin. We had the opportunity to digitize a neolithic burial at the "Landesamt für Denkmalpflege und Archäologie Sachsen-Anhalt" using a combined approach of 3D-scanning and photogrammetry.
Due to the fairly complex geometry of the object, some areas could not be captured by the 3D-scanner. Therefore close-range photogrammetry was employed to fill the gaps. Furthermore it served as basis for texturing the 3D-scan with high quality imagery.
Hope you like it.
Cheers!
here are some results of a small project I did within my seminar "3D-documentation" at the University of Applied Sciences, Berlin. We had the opportunity to digitize a neolithic burial at the "Landesamt für Denkmalpflege und Archäologie Sachsen-Anhalt" using a combined approach of 3D-scanning and photogrammetry.
Due to the fairly complex geometry of the object, some areas could not be captured by the 3D-scanner. Therefore close-range photogrammetry was employed to fill the gaps. Furthermore it served as basis for texturing the 3D-scan with high quality imagery.
Hope you like it.
Cheers!
You do not have the required permissions to view the files attached to this post.
-
- V.I.P Member
- Posts: 904
- Joined: Fri Apr 20, 2012 1:49 pm
- 11
- Full Name: Thad Wester
- Company Details: Clarity Scanning
- Company Position Title: President
- Country: USA
- Linkedin Profile: Yes
- Location: SC
- Has thanked: 1 time
- Been thanked: 2 times
Re: Results of combining 3D-scans and photogrammetry
WOWOWOW. Incredible.
Would love to learn more about your methodology.
Would love to learn more about your methodology.
- sreed
- V.I.P Member
- Posts: 257
- Joined: Mon Apr 28, 2014 4:31 pm
- 9
- Full Name: Scott Reed
- Company Details: Prologue Systems
- Company Position Title: Principal
- Country: USA
- Skype Name: sabotva
- Linkedin Profile: Yes
- Location: Richmond, VA USA
- Has thanked: 16 times
- Been thanked: 15 times
Re: Results of combining 3D-scans and photogrammetry
Yeah! What #tbwester said! I'd like to hear your workflow. Looks impressive. Any way you could share the results through something other than photographs, as beautiful as they are?
- arie
- I have made 40-50 posts
- Posts: 47
- Joined: Wed Jun 05, 2013 11:43 am
- 10
- Full Name: Arie Kai-Browne
- Company Details: University of Applied Sciences Berlin
- Company Position Title: Scientific staff member
- Country: Germany
- Skype Name: aaaarie
- Linkedin Profile: No
- Location: Berlin
- Been thanked: 11 times
Re: Results of combining 3D-scans and photogrammetry
Thad, thanks! Scott, unfortunately I can't share the dataset itself. Sorry.
The workflow consist of following steps:
First, the structured-light scans have to be processed to derive a continous, high-resolution mesh. In our case, we use a breuckmann smartSCAN 5 HE with OPTOCAT but for the methodology it does not matter which structured-light scanner is used.
Parallel to that, a dense pointcloud has to be computed from the close-range photogrammetry using Agisoft Photoscan (but it also has been tested with VisualSFM + Meshlab).
For combining the datasets I use Cloudcompare to transform them into the same coordinate system using reference-markers and/or an ICP-based approach. After that the distance between the two datasets has to be calculated. Areas, which could not be captured with the 3D-scanner, have a higher deviation than the overlapping areas. The photogrammetric pointcloud is segmented based upon the distance value, only leaving points, which are located in the gaps of the 3D-scan. The segmented photogrammetric pointcloud and the 3D-Scan mesh (or rather the points of the mesh) are merged and remeshed using the Poisson reconstruction. The new mesh is reimported to Agisoft Photoscan for texturing, but this too can be done with VisualSFM and Meshlab (or other commercial software).
In this case, the dataset consist of approximately 18 million polygons and a color texture (16384 x 16384 px). The rendering of the images was done in Blender, using the path-tracer Cycles.
Also, I have to add, that I can't emphasize enough how valuable Cloudcompare has become for us! It's become a common tool within my seminar as well as for my research work! So a big thanks to Daniel and all involved in Cloudcompare.
Cheers!
The workflow consist of following steps:
First, the structured-light scans have to be processed to derive a continous, high-resolution mesh. In our case, we use a breuckmann smartSCAN 5 HE with OPTOCAT but for the methodology it does not matter which structured-light scanner is used.
Parallel to that, a dense pointcloud has to be computed from the close-range photogrammetry using Agisoft Photoscan (but it also has been tested with VisualSFM + Meshlab).
For combining the datasets I use Cloudcompare to transform them into the same coordinate system using reference-markers and/or an ICP-based approach. After that the distance between the two datasets has to be calculated. Areas, which could not be captured with the 3D-scanner, have a higher deviation than the overlapping areas. The photogrammetric pointcloud is segmented based upon the distance value, only leaving points, which are located in the gaps of the 3D-scan. The segmented photogrammetric pointcloud and the 3D-Scan mesh (or rather the points of the mesh) are merged and remeshed using the Poisson reconstruction. The new mesh is reimported to Agisoft Photoscan for texturing, but this too can be done with VisualSFM and Meshlab (or other commercial software).
In this case, the dataset consist of approximately 18 million polygons and a color texture (16384 x 16384 px). The rendering of the images was done in Blender, using the path-tracer Cycles.
Also, I have to add, that I can't emphasize enough how valuable Cloudcompare has become for us! It's become a common tool within my seminar as well as for my research work! So a big thanks to Daniel and all involved in Cloudcompare.
Cheers!
You do not have the required permissions to view the files attached to this post.
- 3DForensics
- Honorary Member
- Posts: 1979
- Joined: Mon Aug 03, 2009 1:52 am
- 14
- Full Name: Eugene Liscio
- Company Details: AI2-3D Forensics
- Company Position Title: Owner
- Skype Name: eliscio
- Location: Toronto, Canada
- Has thanked: 13 times
- Been thanked: 70 times
- Contact:
Re: Results of combining 3D-scans and photogrammetry
Talk about coincidence...
I was just coming to the forum to post a link that was in my inbox this morning about scanning in a cave to recover some old Homo naledi fossils.
youtu.be/
Then I saw another video for the same area
youtu.be/
I love how crisp and sharp your data looks. I would like to see some of that data up close. Great Work!
Oh yeah...as for CloudCompare, I can't say enough good about the program and the contribution that Daniel has made to the 3D community.
Eugene
I was just coming to the forum to post a link that was in my inbox this morning about scanning in a cave to recover some old Homo naledi fossils.
youtu.be/
Then I saw another video for the same area
youtu.be/
I love how crisp and sharp your data looks. I would like to see some of that data up close. Great Work!
Oh yeah...as for CloudCompare, I can't say enough good about the program and the contribution that Daniel has made to the 3D community.
Eugene
- arie
- I have made 40-50 posts
- Posts: 47
- Joined: Wed Jun 05, 2013 11:43 am
- 10
- Full Name: Arie Kai-Browne
- Company Details: University of Applied Sciences Berlin
- Company Position Title: Scientific staff member
- Country: Germany
- Skype Name: aaaarie
- Linkedin Profile: No
- Location: Berlin
- Been thanked: 11 times
Re: Results of combining 3D-scans and photogrammetry
Hi Eugene,
I was wondering when the forensics would show up here...
Attached are some crops of the rendered images. Zooming in further than that reveals the fairly heavy decimation and the loss of fine details, which is quite ashame. But handling and rendering 18 million polygons is just so much easier than +100 million...
Cheers!
I was wondering when the forensics would show up here...
Attached are some crops of the rendered images. Zooming in further than that reveals the fairly heavy decimation and the loss of fine details, which is quite ashame. But handling and rendering 18 million polygons is just so much easier than +100 million...
Cheers!
You do not have the required permissions to view the files attached to this post.
- 3DForensics
- Honorary Member
- Posts: 1979
- Joined: Mon Aug 03, 2009 1:52 am
- 14
- Full Name: Eugene Liscio
- Company Details: AI2-3D Forensics
- Company Position Title: Owner
- Skype Name: eliscio
- Location: Toronto, Canada
- Has thanked: 13 times
- Been thanked: 70 times
- Contact:
Re: Results of combining 3D-scans and photogrammetry
Don't worry...we've got people and eyes everywhere!arie wrote:I was wondering when the forensics would show up here...
Seriously, those images are really great. Both from the perspective of quality and content...who doesn't like to see a well preserved skeleton!?!?
Eugene
- gilles_3DR
- V.I.P Member
- Posts: 235
- Joined: Mon Dec 08, 2014 11:41 am
- 9
- Full Name: Gilles Monnier
- Company Details: Hexagon
- Company Position Title: 3DR Technical Manager
- Country: France
- Linkedin Profile: Yes
- Has thanked: 2 times
- Been thanked: 28 times
Re: Results of combining 3D-scans and photogrammetry
Hi arie,
Very very nice model. Thx for sharing. I always like when several data sources can be combined to get the best of each acquisition technic!
But then, finding the right workflow can be quite challenging!
I guess that the part modeled by photogrammetry is less accurate than the ones modeled by 3D scans. Did you get an estimation of the accuracy difference?
Very very nice model. Thx for sharing. I always like when several data sources can be combined to get the best of each acquisition technic!
But then, finding the right workflow can be quite challenging!
I guess that the part modeled by photogrammetry is less accurate than the ones modeled by 3D scans. Did you get an estimation of the accuracy difference?
- arie
- I have made 40-50 posts
- Posts: 47
- Joined: Wed Jun 05, 2013 11:43 am
- 10
- Full Name: Arie Kai-Browne
- Company Details: University of Applied Sciences Berlin
- Company Position Title: Scientific staff member
- Country: Germany
- Skype Name: aaaarie
- Linkedin Profile: No
- Location: Berlin
- Been thanked: 11 times
Re: Results of combining 3D-scans and photogrammetry
Eugene, thanks! It definitely was quite aesthetically pleasing to visualize this dataset.
Gill, due to the methodigical approach of combining both datasets, a deviation map between the two had to be computed anyway- the attached image shows how close these two aqcuistion methods can be. The histogramm has been adjusted slightly, red begins at 1mm deviation.
I should add, that the images were not taken under ideal conditions for a photogrammetric reconstruction; the light source changed positions, which usually should be avoided. But in this case it was done on purpose, since I wanted the color information to be as shadow-free as possible. It's a lot more fun visualizing the dataset.
Cheers!
Gill, due to the methodigical approach of combining both datasets, a deviation map between the two had to be computed anyway- the attached image shows how close these two aqcuistion methods can be. The histogramm has been adjusted slightly, red begins at 1mm deviation.
I should add, that the images were not taken under ideal conditions for a photogrammetric reconstruction; the light source changed positions, which usually should be avoided. But in this case it was done on purpose, since I wanted the color information to be as shadow-free as possible. It's a lot more fun visualizing the dataset.
Cheers!
You do not have the required permissions to view the files attached to this post.
- danielgm
- V.I.P Member
- Posts: 195
- Joined: Thu Jun 20, 2013 10:45 am
- 10
- Full Name: Daniel Girardeau-Montaut
- Company Details: CloudCompare
- Company Position Title: Administrator
- Country: France
- Linkedin Profile: Yes
- Location: Grenoble, France
- Has thanked: 11 times
- Been thanked: 34 times
- Contact:
Re: Results of combining 3D-scans and photogrammetry
I'm glad to serve the community, especially when I see such a gorgeous result!
P.S.: today I just met Marco Callieri from the Visual Computing Lab and he presented the new version of Meshlab that should come out in a few months. It's gonna be great! Hopefully in the future we'll manage to increase the interoperability between both tools.
P.S.: today I just met Marco Callieri from the Visual Computing Lab and he presented the new version of Meshlab that should come out in a few months. It's gonna be great! Hopefully in the future we'll manage to increase the interoperability between both tools.