Thanks for your comment on the Haro3D library. Are you using version 1.0 or 1.2?
This .dll problem generally occurs when the Kinect20.Fusion.dll is not explicitly included with the build. In the "Source Files" category of your Build Properties, add "Kinect20.Fusion.dll" to the "Always Included". Make sure to select the version of "Kinect20.Fusion.dll" that corresponds to your LabVIEW version (32 or 64 bits). There are two .dll's, one for each version, and they have the same name but are not the same file.
You can also fix the problem by manually copying the correct version of Kinect20.Fusion.dll in the same directory than the HaroTek_Kinect2_x32.dll of your executable (typically the \data directory).
Thank You Marc for your reply,
I'm using version 1.0.
Window 10 (64 bits OS)
I have tried both the methods you mentioned but it still give me the same problem.
You should only put Kinect20.Fusion.dll in "Always Included". If you put the whole directory, or if you add the other dll, you will end up with 2 copies of the HaroTek dll, which creates the same dll. problem. The resulting compiled executable cannot be fixed by copying Kinect20.Fusion.dll.
I have an application where I would like to use the Kinect 2 to track just one hand over a keyboard (including joints). Can this library to useful for such application?
The Kinect V2 tracks only the wrist, the hand, the thumb, and the tip of the hand. If you want to track individual fingers, you might want to consider the Leap motion. A library is available from MakerHub.
Tip of the hand might be enough. This application is for a seriously injured veteran who only has limited mobility with one arm and fingers. A VR helmet would not be feasible. I am just using a webcam right now to image the keyboard area to determine hand position (a bright white sticker on the back of his fingernail as a fiducial). I am going to see if the Kinect 2 would be useful to track the hand directly without the use of a fiducial and in a broad range of lighting environments.
As MarcDub said, looks like Leap Motion might suit your application better. Kinect is much better at full body skeleton detection than finger level.
The Kinect has active lighting, so it should work in most indoor ambient light environments (direct sunlight would be an issue). I suppose that accuracy would be the issue. You probably want to keep the Kinect as close as possible of the keyboard. Accuracy decreases with distance. However, the Kinect needs to be far enough so that the human is recognized. Sitting positions are usually OK. The absolute minimum distance is 0.5 m. So the sweet spot is probably around 1.0 m.
Let us know how your project is going by posting updates in the 3D Vision group.
Did I missed out anything ?
I still get the same error message while trying to run the exe.
But it works after i tried one of the method from the ni forum. ( By the specifying the path to the dll from the Call Library Function Node).
Regarding the last sentance of my last post, it work on calling the DLL (no more pop up), but there was an error out from the Call Library Function Node.