Help ! can't create complete mesh.

[agga]

Hi,

I have 59 5MP camera's for a (half) body scan.
Foto settings: f/2.9, ISO 100, 1/10sec.
In the workflow I add the folder. (camera calibration: 0.0014 x 0.0014)
All camera's align ok!

I create a medium dense cloud. (also tried high and ultra high)

And when I then build the mesh I always get missing parts in the mesh... sometimes the hair, sometimes the face, somtimes the back... but always a different part. I let people always face the same side... camera positions are also fixed.

What am I doing wrong ? Why can't I create a nice mesh?
Can someone help me here ? getting desperate...

[Exhale]

Hi ,

Can you upload the photos and pm me the link. So I can test it on my system and let you know about the issue.

regards

[agga]

Hi Exhale,

sended you the link in a pm.
Have a lot of other examples...

[Wishgranter]

Hi Aga, if can send me the link too ( PM ), will digg intro that. what precise camera model you use ? its something from medium format camera ? 

59 x 5 Mpix images. have read it again.... 

[fengkan]

I am justing wondering are you using raspberry pi?

[agga]

yes, raspberry's.

cheap, network attached, easely programmable...

[Tyler J]

What camera are you using (just out of curiosity)

[agga]

the camera is a Omnivision OV5647 (http://www.ovt.com/products/sensor.php?id=66)

[fengkan]

I also tried 20 rpi cameras for head scanning, but the result is not satisfying.
What parameters do you use to take photos?

[agga]

pretty default settings; -t 1 -iso 100 -n

[fengkan]

How is the quality of your photo? I explicitly specified the awb and ex params, but the color of photo still differs from each other. I think this made alignment difficult.

[Alexey Pasumansky]

Hello agga,

If possible, please send the link to the dataset to support@agisoft.ru.

One of the most probable reasons of the problem is insufficient number of matching points found on the face/head. It may be caused by ineffective use of frame space, for example, the face occupies only a small area on the image, or the image is too noisy or is not focused properly on the object.

[Wishgranter]

have done few reconstruction will reply in 48 hours.....

[voudas]

can some1 upload a set so that we can test as well? ty!