[~Cory~]

As always there will be no end all be all answer. You need to make the decision off of your current resources, time frame and abilities.

Remember that successful implementations will need to get the image, filter, detect, find the information about the targets, send them to the robot, the robot needs to receive information, actuate the respective actuators and implement some kind of feedback loop.

Roborealm can do a majority of these tasks with little actual coding and there is some documentation related specifically to FIRST. Roborealm, however needs a somewhat powerful Windows computing environment to be successful.

OpenCV and its forks are awesome, powerful and well documented. They can run on very simple devices and full blown Windows is not needed. We used the AFORGE.NET c# library last year and had to do a lot of heavy lifting, and javaCV will force you do the same.