No announcement yet.

Problems scanning a thin item

  • Filter
  • Time
  • Show
Clear All
new posts

  • Problems scanning a thin item

    I'm having a lot of difficulty getting a good scan of a thin object using 3df Zephyr. I've looked at some of the articles on the forum, such as this one, but I still have been unable to achieve good results:
    I welcome suggestions.

    For this project, we are scanning Viking weapons in order to get detailed measurements of these objects to use for modeling the physics of the weapons when they are wielded.

    I've been able to get excellentd scans of Viking spearheads and axeheads. But a good scan of a Viking sword continues to elude me. The resulting model is good over most of the sword, except at the point, where the object is both thin and small. The two flat sides of the sword are well modeled, but they don't meet along the edges or at the point. It is as if the front and back flats of the sword were two sheets that were imperfectly aligned and then attached to one another - they are offset laterally by a few mm.

    My work flow:
    1. I've created a jig for the sword, so the camera, lights, and contrasting color background are fixed, and I can rotate the sword around its long axis to get pictures around one orbit, then move the sword along its length to get the next orbit, and so on, along the length of the sword.
    2. The historical sword (which I can access for my tests only rarely) has plenty of surface texture and needs no treatment. The modern replica I use as a stand-in is shiny, so I dust it with white powder to dull it, then apply blotches of contrasting color chalk dust to create a random pattern over the surfaces.
    3. I take many photographs, with closely spaced photos at the edges, and more widely spaced photos of the flats. I also shoot oblique shots all around the point. I'm using the Lite version, so am limited to 500 images, and I aim to get that many images of the entire sword.
    4. I mask the images in Masquerade to eliminate the background, leaving only the sword.
    5. I process the images with advanced settings, with all the parameters set to the highest level.

    Generally, all the cameras are accepted except for the edge shots. The sword, on edge, takes only a small part of the camera frame, and I presume there is simply no detail to capture and match with other views. Inspecting the images from the camera, they seem to be sharp and detailed with sufficient depth of field

    I welcome suggestions or recommendations. Thank you.

  • #2
    Hi William R. Short - Thanks for sharing your workflow - I've scanned swords before and they can be tricky!

    Are you able to share a photographic dataset? - Google Drive download link or similar works well for us.

    It's generally not recommended to use the highest settings, especially in the initial camera orientation stage, as this can increase the likelihood of false positives.

    Once I've had a look over the images I'll get back to you with some tips :-)


    • #3
      Originally posted by cam3d View Post
      Are you able to share a photographic dataset?
      Thank you cam3d
      Images and masks are here:

      Originally posted by cam3d View Post
      Once I've had a look over the images I'll get back to you with some tips :-)
      I'd be grateful for advice and suggestions.


      • #4
        Great thanks for sharing

        From what I can see you've got the sword in a horizontal orientation - I'm not sure how heavy it is, but is does run the risk of subject deformation and so I'd advise to orient your sword vertically.

        I was able to get a fairly good reconstruction from these settings:

        ?Click image for larger version

Name:	Capture.png
Views:	112
Size:	24.6 KB
ID:	8626
        Click image for larger version

Name:	Untitled-1.jpg
Views:	94
Size:	127.2 KB
ID:	8627

        - There was a bit of inconsistency in the middle portion of the sword where the loops are a little further apart than at the hilt/tip but overall it held together fairly well.

        - I'd recommend additional shots further away which incorporate more of the subject so it is easier for Zephyr to piece it together (like a map to help solve the jigsaw puzzle)

        - Using a motorized turntable linked with shutter release will make captures like these a lot more uniform and robust.

        - The powder is a bit clumpy - If there was something finer at hand or ideally a spray like AESUB you'd get much less surface 'noise' from the scan prep coating.

        - You're very close to getting superb results! Let me know how you get on and if you have any further questions I'll happily answer them as best I can.


        • #5
          cam3d, thank you for your comments and suggestions.
          Sadly, when I ran the image set with the settings you used, the resulting reconstruction was not usable, and not much different from what I obtained with the more aggressive settings. While the two flats were well reconstructed, none of the edge cameras were accepted, and the two flats were imperfectly joined together. It can be seen clearly in this view of the dense point cloud, near the point of the sword looking at the edge (from the run using the settings you suggested):
          Click image for larger version

Name:	edge_rev4.png
Views:	100
Size:	114.4 KB
ID:	8629
          May I ask, did you get a better reconstruction of the edge? Do you have suggestions on how to photograph the edge so the two flats can be joined more accurately? When I ran the same photograph set with the more robust settings, the two flats joined in a point, but they were offset laterally from one another by a few mm.

          Some of your suggestions are easy to implement, while some are more difficult given the restrictions on what we can safely do with the ancient artifacts, but you have given me some good ideas.
          - Yes, the sword was horizontal when photographed. Our attempts to support the swords vertically made both the photographer and the curator of the collection nervous. The swords, both ancient and modern are robust. While I have not measured it, self-deformation seems unlikely. I can measure to verify.
          - We can make the photo-orbits around the object more uniform.
          - Our attempts at using images from further away as a "road map" of the object didn't seem to help when we tried them before. We will add them back in.
          - The powder coating is necessary for dulling the shiny modern sword we use for our practice runs, but both unnecessary (because of surface corrosion and patina) and not permitted for the ancient swords. If you think the powder coating is affecting our ability to get an accurate model of the edge for the modern sword, we can make changes to how we coat the object in our practice runs.

          Again, my thanks.


          • #6
            Hi again William R. Short - Yes I got a better construction which doesn't have a split. I recommend clearing your temporary directory (Tools>Options>Trash Icon on Temp/Autosave directory) and running the project again with the same settings as suggested.

            No need to worry about the powder coating in this context - Thanks for clarifying :-)

            Click image for larger version

Name:	Untitled-1.jpg
Views:	100
Size:	105.0 KB
ID:	8633


            • #7
              Again, my thanks to you cam3d.

              I have two results to share with you. First, I cleared the temporary directory as you suggested, and ran the model with the suggested settings. I got a reconstruction that was pretty good, with the two sides coming together to form an edge. It is certainly a better model than anything I've been able to achieve previously.
              Question: do you suggest clearing the temporary directory for each new project?

              Second, before I read your reply, I had some ideas of my own for better photographs. To save time, I rephotographed only half the blade down to the point. Two changes: (1) I got the camera as close to the object as I could, so that the edge shots covered much more of the frame; and (2) I powdered the blade slightly differently, which probably didn't make a lot of difference. The resulting model of the half blade was excellent, and most of the edge cameras were accepted. If I can get a model that clean of the entire sword, it will be more than adequate for our research project.
              Click image for larger version

Name:	Image1.png
Views:	94
Size:	160.8 KB
ID:	8635
              The sword has been well-used in my organization's Viking combat research, so a lot of the irregularities seen in the edge of the model are not errors in the model, but instead, they accurately reflect the edge damage and wear of the physical sword.

              Many thanks for your suggestions and comments.


              • #8
                Hi again - I've found that if high settings are used which produce non-ideal results, then settings are rolled back to a less process intensive configuration, this can negatively impact further attempts to generate a model.

                I don't know exactly how it works under the hood, but flushing the system of these holdovers by clearing the temp file makes sense, but shouldn't be required every time. - I'll raise this with the dev team and see if they have any insight as to why this is happening.

                Haha I was wondering about the edge, but I could see from the original photos that it was... well used... Looking forward to seeing more of your digitized weaponry!


                • #9
                  I'd love to see some images of the finished sword model when you're done William. I'm a big fan of medieval and Japanese weapons.

                  And thanks for the tip about clearing your temporary directory, Cam. I'll need to make sure I do that from time to time as well.


                  • #10
                    Originally posted by DMG View Post
                    I'd love to see some images of the finished sword model
                    Our plan is to make the models of the historic weapons we scan publicly available. Thank you for your interest.


                    • #11
                      Originally posted by William R. Short View Post

                      Our plan is to make the models of the historic weapons we scan publicly available.
                      That's awesome, Thank you. I look forward to seeing your work when it's complete.