FPGA Implementation of Global Vision for Robot Soccer as a Smart Camera

An FPGA-based smart camera is being investigated to improve the processing speed and latency of image processing in a robot soccer environment. By moving the processing to a hardware environment, latency is reduced and the frame rate increased by processing the data as it is steamed from the camera. The algorithm used to track and recognise robots consists of a pipeline of separate processing blocks linked together by synchronisation signals. Processing of the robots location and orientation starts while the image is being captured so that all the robot data is available before the image has been fully captured. The latency of the current implementation is 4 rows, with the algorithm fitting onto a small FPGA.