How to build a complete (360 degree) model of an object

[swhite]

I have a skull and I am attempting to make a 360° model. I am following the recently updated tutorial "How to build a complete (360 degree) model of an object".  I have made 252 photos of the skull on a turntable, 108 of the top and 244 of the bottom. I imported all photos into one chunk and assigned each photo into a top or bottom camera group. All photos were carefully masked to exclude the background. Next I attempted to align the photos using the suggested parameters (Accuracy high, Generic presentation Estimated, Key points limit 40,000, Tie point limit 4,000, and Apply mask to key points. The result is that all of the bottom cameras aligned while none of the top cameras aligned. When inspecting photos from the bottom (successfully aligned) group the skull is covered in blue dots while in the top group (where none aligned) the skull is covered only in white dots. When I put the top photos in a separate chunk they all align beautifully but not while the bottom photos are present.

My question is how to get all images to align in one chunk. Any comments or suggestions on how to proceed would be greatly appreciated.

[Alexey Pasumansky]

Hello swhite,

May I ask you to send the project saved in PSZ format with the results of the alignment of all the images in the single chunk, so that we could try to check if there's any problem?

[swhite]

Thanks Alexey for looking into this matter.

The PSX file and screenshot of the workspace are attached.

Do you need the files folder as well?

[swhite]

Sorry - I just sent you a link to the PSZ file as well.

[Alexey Pasumansky]

Hello swhite,

I have received the project file from you, but it appears that the are no detected tie points between the Top and Bottom subsets.

According to the image thumbnails, I see that images DSC_3135 - DSC_3170 (the sub-set taken from the side) have been taken while the skull hs been put on its top. Probably adding one more sub-set of images from the similar camera position but with the skull put on its bottom would help to combine both parts of the complete project.

You can also try to run the image matching again for the complete dataset with the Generic preselection disabled, maybe it could help to detect some tie points between the sub-sets.

[swhite]

Good news! I unselected Generic preselection as you suggested and this time all 252 cameras aligned. Am now carrying on with building a dense point cloud.

Thanks

[swhite]

I have now completed the model all the way through and the result is superb!

Many thanks.