A RaspberryPi running a python script was used with a Microsoft LifeCam HD-3000 for vision processing. OpenCV was the vision processing library, and WPILib's network tables were used to pass data from the Pi to the RoboRIO. The post-processing code to analyse the data that was passed to the RoboRIO can be seen here, and the rest of the robot code can be seen here.
This script is called before vis_calibrator.py and vis_rect.py runs to set the camera to the proper settings. For competition, a startup script was made on the Pi that ran this script before it ran vis_rect.py.
This script is use to calibrate the vis_rect.py, the main vision processing script, to the hsv threshold of the target. After a calibration, the hsv values were updated in vis_rect.py. In order to view the script's GUI, the script was ran on a laptop that was connected to the camera. The laptop also had Ubuntu, OpenCV, and other necessary libraries to simulate the Pi and calibrate the main script.
This is the main vision processing script that outputs arrays of x and y coordinates, widths, and heights of the seen targets. In this case, each rectangle is viewed as a separate target. The arrays are passed to network tables hosted by the roboRIO.
Installing OpenCV on a RaspberryPi: http://www.pyimagesearch.com/2016/10/24/ubuntu-16-04-how-to-install-opencv/
OpenCV and Python tutorials and documentation: http://opencv-python-tutroals.readthedocs.io/en/latest/index.html