Recording video to HD

[appinheiro]

Dear,

I have some questions about camera record. They are:

1) How I can do to record the video (captured by a FLEX100) to a avi file in computer Hard Disk ?

2) Is there a routine in SDK to record video in a file ?

3) How many cameras I can use into a same PC ?

thanks in advance and my apology for my english

[VincentG]

1 - I'm currently not aware of any way to record the gray scale images to a video feed file.

2 - Not that I know of, but I will defer to the development team........

3 - When we release the latest version of the software, there should be support for up to 24 cameras.....

[beckdo]

Here's a bit more information:

You can't record video from the OptiTrack SDK/API. There are a couple options:

OptiTrack DirectShow Filter (available in the downloads section). From this you should be able to find a suitable program that will convert a DirectShow video stream to an AVI.

If you attempt to use the OptiTrack SDK directly, you can use the GetFrameImage on each NPCameraFrame to get the grayscale images. As you fetch the images, you would likely either push them out to separate files or to something that can generate a video file (ffmpeg for example).

[appinheiro]

If you dont have a way to record in a file the cameras images, i think that the coordinates of the markers are processed in real time. But, imagine the case where the program has a lot of tasks forbidding the real time processing. In this case, is needful record the video camera in HD and later, process the video (avi file).

[Birch]

When the cameras are identifying markers and their coordinates, the processing does not occur in the PC, it occurs inside the cameras. This is the advantage of using OptiTrack cameras, they remove the burden of processing real-time video from the PC and offload it into the cameras. The task of the PC is then limited to collecting the stream of already processed 2d marker data from the cameras.

[leith]

There are some hybrid solutions I've seen in other computer vision cameras, where the majority of the frame is tossed away and only pixels in areas of detail are sent, capping the number of pixels that will be sent per frame to be safe. I think the theory is that a more exhaustive sub-pixel fit of the details can be done later in post, replacing the faster calculation done in realtime. Also, parameters having to do with the centroid can be adjusted in post. In practice though, I've found its pretty rare your markers are big enough in frame that sub pixel accuracy amounts to much. And its likewise rare that you have enough resolution that adjusting eccentricity thresholds really amounts to much.

[appinheiro]

I could understand now, thanks for your attention.

But I have other question - the last
In motion capture system that works on indoor environments, the tecnique described by you does work very well. But, in outdoor environments, where the contrast between the markers and background is little, I image that a solution is other.
In the last case, is need show full image and use algorithns for circle fitting. Do you have a solution ou idea which camera type i could use ?

Other question: Do you sell raw material (reflective strip) bigger than 1"x 12" ? If yes, how much (size and price) ?

[Jim]

Hello:

Our cameras use a Global Shutter with an illumination strobe, so the signal to noise ratio is very high and can work with sunlight, but not in direct sunlight. You can optimize the settings for the environment via software.

Of course, you can make your own algorithms, using our SDK. I would recommend the V100 camera, as that will get you all the options you want.

We do sell bulk reflective material, but you should email sales@naturalpoint.com for a quote. Just ask them for how much you need. Usually we have it by the 1" wide strip, but it can be as long as you want. When you buy it online, if you ordered 10', it would be one long strip, not in 1' sections.

[leith]

I'm curious. Are you referring to human intervention? Or a different approach to circle and blob fitting than is typical of reflective mocap? Most approaches to optical mocap in the visible range center on pattern recognition, 2d bar coding and projection of rectangular objects in 3d space.