LabVIEW
OpenCV on NI1752 Smart Camera - VxWorks cross compilation
Solved!
While that discussion you link to is interesting, it is pretty outdated (2008). At least in respect to ARM support, OpenCV officially has an OpenCV4Android SDK with an armv7a architecture. So for the newer ARM based cRIO it should be possible to recompile the OpenCV libraries in the NI provided Eclipse IDE. And of course the same applies even more for the x86_64 based high end cRIOs.
A recompilation with the NI Linux Realtime specifc Eclipse toolchain is most likely unavoidable and there are most likely some additional config changes that need to be done to the make/configure scripts to make it all work.
You are in so far correct that porting of the OpenCV library to the PowerPC/VxWorks based cRIOs is most likely a herculean undertaking with little prospect for success.
It's completely unimportant where you store the DLL. The VxWorks based OS on the Smart Camera doesn't understand it and can't load it into memory. The DLL is a Windows specific executable format that only Windows knows how to properly load into memory to execute it. Even more so, Windows DLLs are binary executable code modules compiled for the Intel x86 processor. Even if your VxWorks system would know how to load the DLL into memory, its contents would be nothing more than unintelligent gibberish for the PowerPC processor in your smart camera.
It's akin to trying to load a JPG image in an application that only understands Windows bitmaps: The JPG image only contains random binary data for that application.
If you want to use OpenCV rather than IMAQ Vision on this smart camera, your only option is to have another program on a Windows PC which does your image analysis and then from your smart camera send the acquired images to that application (over direct TCP/IP communication or over a shared network drive as file) and eventually send the processed data back afterwards.
Another option would be to maybe use a different smart camera. NI has the NI 1772, NI 1772C, NI 1774, NI 1774C, NI1776, NI 1776C, NI 1778 line which use an Intel compatible processor and the Pharlap ETS OS, which with several caveats can load Windows compatible DLLs. One of the caveats is that the DLL and all its dependencies need to be compiled with one of the supported Visual C compilers, VC 6, VC 2003, VC 2008, or VC 2010. Any other compiler including MinGW and other Gnu C based compilers will generally not work unless you are a real Gnu C linker guru and know exactly what command lines to tweak to get the right binary.
There might be new NI Linux Realtime based smart cameras in the making, but I don't know. They would be likely the most trivial to get OpenCV running on, as the user space environment is very much Linux like.
That obviously can't work. You need to write two independent applications. One for your Windows computer which receives the image through TCP/IP communication or through a shared network location as a file and calls your OpenCV DLL to do your analysis and another one that does the image acquisition on your smart camera and sends the images to the Windows application.
Maybe there is a possibility to access your smart camera directly from your Windows application by installing some drivers in your Windows system but I have no experience with the NI smart cameras, so can't tell you what can and cannot be done with them.
In any case you have to put the application for your Windows system under the My Computer group in your project, not under the NI1752 Smart Camera group for your device. If the smart camera can be accessed as a simple camera under Windows your program you show in the image should basically work, provided you can see the camera in NI MAX under "My System->Devices and Interfaces" and use the device name assigned there instead of img0.
This is explicitly not the name as shown in NI MAX under "Remote Systems".
