Robot Identification using Shape Features on an FPGA-Based Smart Camera

An FPGA based smart camera captures images of soccer robots, and processes them using stream processing to identify the individual robots. Three standard methods of distinguishing the robots based on shape features are compared, and the advantages and disadvantages of the different methods discussed. Compactness and moment based measures give poor performance because of the sensitivity of the low resolution shapes to relatively small distortions. Normalised area gave the best results in this application.