Design Overview
The main idea is to build a vision system in hardware, for RoboCup. A colour CCD camera is used for input, which is then decoded into digital video before being sent to the Field Programmable Gate Array (FPGA). We have been using the Altera Stratix PCI development board, but may swap to a Cyclone II development board soon.
The final FPGA design will process the digital data to produce coordinates of detected objects. These coordinates can then be passed onto the rest of the robot.
The design performs colour-based object recognition. In RoboCup, each object of interest is a set colour. Outputting coordinates for each coloured object detected will allow robots to quickly decide the best action to take.
It is intended that each frame can be processed in real time. Currently, thresholding and labelling is performed on each pixel in real time. Object detection is still only implemented in software simulations. However, the results so far look promising.