Design Of Humanoid Soccer Robot Decision-making System Based On Vision

Humanoid soccer robot game, which has a significant relationship with the fields ofintelligent control, pattern recognition, signal processing, sensor technology and wirelesscommunication technology, is a hot topic in recent years. The humanoid soccer robot visionsystem and decision-making system are deeply researched with the methods of experiments,mathematical modeling and computer simulations.Aiming at the shortcomings of recognition ability in a complex field environment, ahumanoid soccer robot vision recognition system based on ARM is designed. In terms ofhardware design, Samsung S3C6410is chosen as the main image processing chip, FreescaleMC9S12XS128microcontroller is applied as the servo controller, and Logitech C270camerais used as the visual sensor. In aspect of software design, embedded Linux platform isestablished, and OpenCV image processing libraries are transplanted to realize the function ofcapturing surrounding pictures in the game.A mixed segmentation algorithm, which combines the threshold algorithm and regiongrowing algorithm, is used to segment the color information rapidly. Centroid method andcircular degree detection algorithm are used to recognize the target. It is verified that thisdesign can be used to identify the color swatch and locate the center coordinate in a complexenvironment accurately.A physical model and a mathematical model are established, which is on the camera andmonocular vision measuring system. A calibration experiment is made to obtain the camera’sinternal parameters. The feasibility of the visual measuring system is analyzed theoretically.The method combing the forward vision and electronic compass is adopted, which can beused to position the robot.A fuzzy decision-making system is designed. First, basic movement commands, whichare needed in the robot game, are designed to be the strategy base. Second, deviationsbetween the expected value and the target center coordinates after image processing are takenas the inputs. Basic motion command numbers are taken as the outputs. It is verified thatmotion command numbers can be accurately outputted when the inputs are certain settingvalues.