I wish to use PCL on a project that uses a Odroid X2. Will this be problematic? I'm going to start by making some tests on a different machine, but I wonder what will happen once I transfer the project over to the Odroid.
Some time ago I compiled PCL on an Odroid U2.
Performance is really decent, but expect problems with OpenGL etc..
Graphics drivers are always a problem on these platforms because the lack of openness of the hardware.
All the computations will eventually slow down but you can still use OpenMP. You would have to benchmark specific code to get numbers of how the performance is affected.
I don't want to use any visualization on the Odroid since it's the processor for an aerial vehicle (which has no screens). The only thing I need is for it to run the node I'm going to create. I think I will have to test it to understand the problems I might be getting myself into.
On the original topic:
Is it possible to have a cloud structure in the node and keep adding more points to it (as the target detection nodes keeps outputing them)? I ask this since I was thinking about keeping a two copies of the cloud (original and filtered) and keep adding points to the original one and filtering it (as well as using some kind of clustering strategy) to get an estimate on the position of the targets. Is this a good strategy? Sorry, but it is my first time dealing with this kind of problem.