Problem with sphere detection

[krl]

Hello, I am relatively new to RTI, but have successfully used the software a few times. I keep getting stuck on the sphere detection step of my latest project, which is a highlight-based PTM fitter. I crop out the spheres, hit sphere detection, but once it's done the next button is still not click-able. I have deleted the project and restarted several times to no avail. I was able to make a highlight-based HSH fitter with the same piece. What am I doing wrong

[Dave Martin]

Krl,
Can you post one of your source images with the spheres, and then a screenshot of where you say you "crop out the spheres"?
Dave

[krl]

https://imgur.com/a/cYAXSdh

https://imgur.com/a/ticNYo2

It looks like there's a 0.5 MB cap to these posts, but here are links to the images you requested.

Thanks a lot for your help!

[Dave Martin]

Thanks Krl,

I was worried when you said you were 'cropping' them - I thought you were cropping them out of the image before processing! - those boxes are just to define the search area.

Secondly, just looking at your original image, I have a few observations of which only the first couple may relate to the problem you have experienced.
1. The background on which the spheres are sitting (or is behind them) looks very close in colour to the spheres, because you're using a photograph where the grey card or background is in shadow - you're not doing the circle-edge detection any favours, its trying to detect black on near-black!
2. The illumination in the shot you kindly supplied looks rather low - the recommended light positions are no lower than 15 degrees to the plane of the object, and no higher than 65 degrees to the plane of the object.
3. The highlight-detection method of RTI generation depends on accurate detection of the highlight on the sphere as each frame is processed, and using that to deduce the direction from which each frame was illuminated. The light should be reflected as a 'spot', as near to a point - but the light looks to be more like patches on your spheres? (maybe either over-illuminated or something like a flood-light used rather than as close as you can mange to a point source). 'Wide' sources won't affect the sphere detection but it will give less-distinct shadows and hence degrade the final result.
4. The whole process depends on each frame having, as close as you can, only one direction of illumination. Looking at your sample image, it looks as if the are other light sources or something out of shot that are throwing multiple direction illumination on the spheres? - that shouldn't affect the feature detection but it will degrade the final result.

I'd highly recommend that you go through the RTI Capture Guide which CHI publish at http://culturalheritageimaging.org/What_We_Offer/Downloads/RTI_Hlt_Capture_Guide_v2_0.pdf - have a look at how the spheres and illumination appear in successful use. CHI also have tutorial data sets on their web site which are known to work, as well general practise they allow you to see how successful capture looks.

cheers/Dave

[krl]

Thanks a lot. I was able to figure it out.

[Dave Martin]

Thanks Krl,

Especially to help anyone who might have similar problems in future, could you share what you did to make it work?

cheers/Dave

[krl]

I reinstalled the software and suddenly it worked. Sorry that probably isn't very helpful.

Thanks again for your advice on the setup!

[Dave Martin]

Thanks Krl,

Main thing is that you're working but it might help someone else in the future!

cheers/Dave