About CModuleVectorProcessing Class

Post Reply
symarobofr
Posts: 8
Joined: Fri Jul 08, 2011 2:16 am

About CModuleVectorProcessing Class

Post by symarobofr » Tue Jul 12, 2011 5:01 am

Hi,

I am building an application to track the position of a camera with respect to 4 markers.

I utilise some code very similar to the code in the Vector Tracking sample. Can you explain me about the data returned by the GetResult, GetPosition and getOrientation functions of the cModuleVectorProcessing object? These positions and orientations are with reference to which co-ordinate frame?
Can i use this class with all camera modes? Or is it restricted to the Segment mode as in your Vector Tracking sample?

I want to use 4 or more retroreflective markers mounted on a marker base (also from NaturalPoint) as the target to be tracked. How do I go about specifying the 3D co-ordinates of these 4 markers on the base to my application?

Does the cModuleVectorProcessing class work with more than 1 marker? If yes, what are the conditions for size, colour, etc. for choosing and placing them? how do you uniquely identify them?

Thanks in advance for your help.

- symparobofr

beckdo
NaturalPoint Employee
NaturalPoint Employee
Posts: 519
Joined: Tue Jan 02, 2007 2:02 pm

Re: About CModuleVectorProcessing Class

Post by beckdo » Wed Jul 13, 2011 12:16 pm

The cModuleVectorProcessing object's results are returned with respect to the camera being at the origin looking down the Z axis. So as you move the vector clip towards the camera, the positional Z value approaches zero.

The class works with all tracking modes, so Segment, Object, and Precision mode.

The cModuleVectorProcessing object is designed to track only NaturalPoint's vector clip and Track Clip Pro. It's a single camera algorithm.

If you want to track a different arrangement or marker count you'll need to roll your own single or multi-camera tracking algorithm or step up to the Tracking Tools & Tracking Tools API which allows you to track arbitrary arrangments and arbitrary marker counts using multiple cameras.

D

symarobofr
Posts: 8
Joined: Fri Jul 08, 2011 2:16 am

Re: About CModuleVectorProcessing Class

Post by symarobofr » Wed Jul 20, 2011 10:25 am

Hi Doug,

Thanks for your reply. The point is i already developed my single camera algorithm.

First,
I just want to know how to enter the 3D co-ordinates of the markers fixed to the markerbase. (We bought Part number MCP1145 found at your website link http://www.naturalpoint.com/optitrack/p ... s-markers/)

I understand I can assume z = 0 and use a ruler to measure the x and y from the centre. But may be thought you might have more precise values since the marker base comes from NAtural point..Any pointers?

beckdo
NaturalPoint Employee
NaturalPoint Employee
Posts: 519
Joined: Tue Jan 02, 2007 2:02 pm

Re: About CModuleVectorProcessing Class

Post by beckdo » Wed Jul 20, 2011 11:01 am

If you have your own algorithm you won't be using cModuleVector or cModuleVectorProcessing as those classes are designed to only track the vector clip or the track clip pro.

Beyond that the Camera SDK only outputs 2D data from the cameras, which would presumably go into your algorithm. Any marker measurements would go directly into your algorithm as well as per your design.

LinusA
Posts: 37
Joined: Mon Nov 08, 2010 8:43 am
Location: Aachen, Germany
Contact:

Re: About CModuleVectorProcessing Class

Post by LinusA » Wed Jul 20, 2011 5:32 pm

Doug,

I guess what symparobofr is trying to say: Given this marker base
Image,
what are the real 3D positions of the marker centers, when you put markers on the studs? I.e. do you have the exact data you produce this marker base from, or did you measure it in a high quality tracking environment and published it? (Actually, as this marker base is pretty much on one plane, the 2D dimensions would be enough).

I was wondering the same, but for my purposes, the data coming out from my algorithm is good enough.

I think what would be convenient is something like this VRML model file for this marker base:
Image

symarobofr
Posts: 8
Joined: Fri Jul 08, 2011 2:16 am

Re: About CModuleVectorProcessing Class

Post by symarobofr » Thu Jul 21, 2011 11:19 am

Doug: Sorry, I am neither using the cModuleVector in my algorithm nor is any 2D data from the Camera SDK going to my algorithm. I just wanted to know the 3D coordinates when i mount the markers on the studs?

LinusA has beautifully understood what i want to know..

LinusA: man you nailed it.. Thank you so much. Thats exactly what i want to know :-)

symarobofr
Posts: 8
Joined: Fri Jul 08, 2011 2:16 am

Re: About CModuleVectorProcessing Class

Post by symarobofr » Fri Sep 23, 2011 2:58 am

Hi Doug,

As LinusA had pointed out, could you please throw some light how to enter the co-ordinates of this marker base? Do you have a VRML file or something? will be of great help. We are losing precision in our application because of this. Thanks for your help.

Post Reply