Forum

Author Topic: Automated masking issue  (Read 3081 times)

DCK

  • Newbie
  • *
  • Posts: 46
    • View Profile
Automated masking issue
« on: August 19, 2014, 09:31:17 PM »
Hello everyone,

I'm having a problem with automated masking that I didn't have with older (>1 year older) versions of PhotoScan. I'm not sure if it's something in the software, or something I'm forgetting about stepping through the process. The attached image pretty much explains the problem. When I use the mask from one image to mask additional images, the imported masks hold onto the shape outline of the original mask. As you can see, the masking of the object (a monkey hip bone) in the additional images is pretty good. I can play with tolerances to find a useful automated mask. However, the retention of the shape from the original mask makes these automated masks useless.

Here's my process.
1. Pictures taken with specimen on turntable.
2. Mask one image.
3. Export mask (either for active chunk or for current photo... makes no difference in end result)
4. Import mask for active chunk.

Any help is much appreciated.

DCK

David Cockey

  • Full Member
  • ***
  • Posts: 207
    • View Profile
Re: Automated masking issue
« Reply #1 on: August 19, 2014, 11:05:47 PM »
Is the object on a turntable with a fixed camera location relative to the background, or is the object fixed relative to the background and the camera moving?

If the object is on a turntable with a fixed camera location relative to the background then there is a process to generate masks starting with a photo of the background without the object. I've never used that process so I don't know the details but it appears to be as simple. Try right clicking on a photo, then select "From Background" as the method. Click OK and you will be asked to select the background photo without the object.

If the object is fixed relative to the background and the camera moving then try this process:

First build a preliminary model:
Align the photos without masking. Low or medium accuracy should be sufficient.
Build the Dense Cloud. Low or medium quality should be sufficient. Trim the dense point cloud to eliminate points from areas you want to be masked.
Build the mesh.

Second create masks using the preliminary model:
Right click a photo.
Click on "Import Mask".
Select "Active chunk" and "From model".
Click on "OK" and masks will be generated.

Now you can align the masked photos, generate a new dense cloud, etc.

« Last Edit: August 19, 2014, 11:08:44 PM by David Cockey »

DCK

  • Newbie
  • *
  • Posts: 46
    • View Profile
Re: Automated masking issue
« Reply #2 on: August 19, 2014, 11:46:31 PM »
Photograph background without object!!! That's it. I knew I was forgetting something.

Life saver. Thank you.