Measurement And Calculation Of The Rotation Of Table Tennis Based On Artificial Intelligence

Artificial intelligence is one of the hottest topics in modern science and technology and it has a wide range of applications in all the fields.As a significant representation of artificial intelligence products,table tennis robot has been developed for more than 30 years.The table tennis robot has already finished the task of hitting low-speed,non-rotating ball,but it can not hit the ball with rotation.There are still some difficulties in calculating the rotation of ping-pong ball,which are also the main study of this article.Table tennis has five core elements: speed,strength,rotation,arc and falling point,of which the most central core is rotation.This paper mainly puts forward some new ideas to calculate the speed,direction and type of rotation based on artificial intelligence algorithm,and the core idea of measuring rotation can be divided into three categories: direct measurement,calculate rotation using trajectory method,and predict rotation with motion method.In this paper,we mainly measure the rotation according to these three kinds of core idea and solve some difficult problems encountered in the calculation process.Finally we verify it by experiments.Using cameras directly shoot table tennis video and calculate the direction and speed of the rotation with trademark on the ball,which is called direct measurement.Multi-camera can not shoot the trademark in every frame because there is only one trademark on one ball.This article designed a algorithm based on monocular camera to calculate rotation speed of table tennis,which mainly uses the displacement trajectory of trademark in 4 consecutive frames or 5 random images.Calculate rotation using trajectory method refers to calculating the direction and speed of the rotation according to the trajectory.Each speed and rotation of the ball will produce a corresponding movement trajectory.Therefore,this paper first obtains some accurate velocity,rotation and falling point data.The determination of motion trajectory is represented by the falling point data.Through the fitting method,it is confirmed that there is a great correlation between these data.That is to say,it is feasible to use the trajectory to calculate the rotation.Then the model of the relationship between motion trajectory and rotation is established through the force analysis of table tennis in three-dimensional space.And the coefficient matrix is calculated through a large number of experiments,which finally determines the specific calculation formula of trajectory backstepping rotation.Simultaneously,calculate rotation using trajectory method requires the data of table tennis trajectory.This paper presents two different algorithms to achieve the task of real-time tracking of table tennis ball.One of the algorithms is based on high-speed photography,and we identify the ball using five characteristics of table tennis.The other algorithm is based on low-speed photography,which is difficult to form a complete and stable feature of the ball.Therefore,we present a new method including the image segmentation and classifier training,then we obtain the highest similarity area that can be marked as the ball.In order to reduce the computational complexity,shorten the recognition time and improve the recognition accuracy,an algorithm for planning the moving ROI region was designed.Finally,the algorithm was verified through experiments.Predict rotation with motion method mainly uses the athlete's idea of predicting rotation of table tennis.The athletes can predict the type and intensity of rotation by opponents' action,so as to lay a foundation for further judging the rotation speed and direction.In this paper,athlete motion data are collected by kinect depth camera,and the improved SVM algorithm is used to classify the motions of athletes.Then the speed of athletes batting is classified by dynamic time warping algorithm.These results can be used to classify the type and intensity of rotation,which can be used to improve the accuracy for the subsequent rotation identification.In this paper,we mainly use the above three methods to calculate the rotation of table tennis based on artificial intelligence.The main application is the visual system of table tennis robot,which lays the foundation for robot to hit and rotate table tennis.The methods can also be applied to the official table tennis competition,and it can real-time display speed of table tennis rotation.This application has two benefits: on the one hand it can give a feedback to the athletes,which will help athletes do more targeted training;on the other hand,it can give viewers a quantitative data to improve the fun of viewing competition,and it has a significant role in the promotion of table tennis.