Hi,

I have an issue with a three original Kinect setup, iPi studio 3.2.2.189 and iPi recorder 3.1.2.36. The sensors are exactly positioned like the example of user manual, with 120 degrees of rotation, giving a triangle of 6 meters sides.
Made many tests, with one minute of perfectly recorded maglite, in different light environments (day and night). I always got a "marker cloud min size error", that always stays at 0. Other values are correct.
Any idea of the problem ?
Thanks

Eric

Hello, Eric

Does the marker gets detected at all?
There should be square visible around the marker on video during the calibration, and 3d points along its way after the calibration has completed.
Please provide some screenshots or calibration project with the video.

What is your graphics card?

-Markers are well detected.
-The graphic card is an brand new Nvidia Geforce GT 730 with 2 go gddr5.
-the third sensor is a Primesense Carmin.
-We have good frames count of 887, average position error 0.002, marker cloud min size 0.485 m (failed), occlusion percent 41 (good).
-When exported to an action scene, cameras are all messed up.

This problem was occuring with iPiStudio 2, but with manual parameters we could rotate manually cameras to put in right position, then recalculate and get good calibration. Now we have no options.

In fact, our studio has a sunny environment, with 12 big windows without curtains, so we must calibrate by night. Do you plan to add back the cardboard calibration feature like in V2 ?

You've said previously that the cloud size is zero. 0.485m makes more sense. That means the range of your motions during calibration is too small. For good calibration, you should move the marker throughout the whole capture area, which is typically 2x2m for depth sensors. So the cloud size would be about 2m.

Occlusion of 40 percent is also far from ideal. Try to make the light visible to all 3 cameras most of the time.


You still can manually adjust camera positions with the controls on the Scene tab.


Calibration with the board is still supported. But technically, it is available for configurations with 2 depth sensors only.

'You've said previously that the cloud size is zero. 0.485m makes more sense. That means the range of your motions during calibration is too small. For good calibration, you should move the marker throughout the whole capture area, which is typically 2x2m for depth sensors. So the cloud size would be about 2m.
Occlusion of 40 percent is also far from ideal. Try to make the light visible to all 3 cameras most of the time.'[/i]
[/color]
At first we got 0, then last trial we got 0.485. But these values are false, because we followed the setup guide rules, which means that we have used the maximum surface for the cloud -more than two meters square.

I suspect that the Carmin sensor we are using (which has a different working area than the regular kinect) is not compatible and put the mess in the calibration process. This has to be said.
We are going to capture with 2 sensors and maybe later buy a third kinect.

Thank you very much for your help.

For calibration, there should be no difference between Carmine and Kinects. Could you please share with us your calibration iPiVideo file, in order we could track down the issue. You can send the file sharing link to our tech support e-mail.

_________________
iPi Soft