Digital Curling Rink System

The Winter Olympics and Winter Paralympics to be held in Beijing in 2022 are major iconic sports events held in my country.With the opportunity of the Winter Olympics,this topic takes the popular sport curling in sports competitions as the background,combines artificial intelligence technology,combines computer vision and robot control,and builds a digital curling rink system.The system consists of a curling observation robot,a curling throwing robot and a human-computer interaction main control terminal.The system can not only simulate and generate throwing strategies for the curling arena to assist users in decision-making,but also allow users to operate the curling robot to perform curling competitions,and even go down to the arena to compete with the curling robot to play an auxiliary training role.First of all,this paper introduces the composition of the curling observation robot,which consists of a vision system,a camera pose control system,a central control unit,an infinite communication module and a motion mechanism.The main task of the observation robot is to collect images in the base camp area and complete the function of scene perception.The most important thing in scene perception is image processing,so this paper firstly corrects the camera distortion for the vision system carried by the observation robot.The vision system plays a key positioning role in observing the movement of the robot,so the camera installation and calibration are carried out.For robot movement control,this paper proposes a motion control method based on visual servoing.Secondly,in-depth research is carried out on the visual detection method of curling robot,and a target recognition and measurement method based on mobile monocular vision is proposed,which realizes the real-time detection and precise positioning of the curling movement process.Specifically,the curling target detection model based on YOLOv5 is implemented.After testing,it can complete the curling detection task of the subject.At the same time,ResNet50 is used as the backbone feature extraction network to correct the coordinates of the curling center point.After experiments,the model reduces the error to about 20% of the uncorrected value.Considering the application scenarios of this system and the mobility and convenience of the observation robot,a method for calculating curling coordinates based on projective transformation is proposed.Finally,an in-depth study of curling sports monitoring and tactical drill requirements was conducted,and a distributed intelligent curling software system was designed.On the basis of the previous research,the curling throwing robot and strategy generation module developed by the research group are integrated with the observation robot studied in this paper,and a distributed computing platform for digital curling rink is developed.The platform can not only simulate curling games,but also generate strategies to display simulations to users,assist users in decisionmaking,and play an auxiliary training role.At the same time,a user operation interface is provided.Users can observe the field score zone information in real time and control the throwing robot to implement the throwing strategy through the human-computer interface,realizing a real curling movement for the public.