Research On Intelligent Pickup Tennis System Based On Visual Analysis

After tennis game is over,it is often tennis balls scattered all over the ground and often manually picked up.Recently,there are almost no intelligent tennis products in markets.The fundamental reason is that efficiency of picking up balls is not high.There is still a lot of work to be done to improve efficiency of picking up balls by robots.This project has developed an intelligent pick-up tennis robot in five aspects:ball-picking robot structure,robot system architecture,robot position positioning and tennis position recognition,ball-picking path planning algorithm,and ball-picking algorithm.A set of intelligent ball picking robot system has been designed,and the following research results have been obtained and have been applied to the National Patent Office for invention patents and utility model patents.(1)Based on the complicated problem of the existing ball picking robot,ball picking mechanism.This project has designed the robot structure as a roller structure.When the robot has rolled over the tennis ball,the tennis ball can be compressed into the collection cage by the weight of the robot's own metal material.(2)The robot system architecture is composed with a host computer,a main camera,and a robot.The main camera is mainly responsible for collecting real-time picture data of the tennis court and sending it to the host computer through WIFI.The host computer processes and extracts the received image data,tennis position and robot position.At the same time,the host computer plans the path of picking ball according to the tennis position.After the planning is completed,the robot sends control commands to the robot.The robot moves according to the control commands received in real time to realize of picking up tennis ball.(3)In order to locate the position and direction of the robot,install a small red block and a large red block at the front and back of the robot,taking the position of the external main camera as the origin of coordinates,establish basic coordinate system OXY,the position and direction of the robot are determined by the positional relationship of the two red blocks.Based on the difficulty of tennis recognition caused by different light intensity and angle of the tennis ball in the tennis court.This project used HSV color space for color experiment analysis and sets HSV color component threshold to determine the tennis position and robot orientation.(4)Used the simple and fast greedy algorithm,the grid method can divide complex problems into sub-module processing characteristics.This project provides a pick-up path planning algorithm based on fusion of grid and greedy algorithm.This algorithm,firstanalyzes the tennis coordinate points,perform rasterization,and then use the greedy algorithm in each divided grid to plan the tennis path from bottom to top.Finally,after the path planning in all grids are completed,connect the grids one by one to obtain the total path.(5)Regarding on the problem of the rotation angle of the robot in the process of tennis picking,the rotation angle is an important factor for picking the next tennis target.This project provided an improved ant colony algorithm for picking the ball path planning algorithm,added rotation angle information to the transfer probability of the original ant colony algorithm.Results of the experiments showed that the improved ant colony algorithm has better effects on the tennis path planning test.(6)The ball picking algorithm is divided into two methods: single camera method and dual camera method.The single-camera ball picking method reads the main camera image data through WIFI,and obtain the robot position and tennis position in the tennis court through analysis and calculation.According to the obtained position information,the robot is controlled to pick up the tennis ball,the advantage is that the picking speed is fast and the disadvantage is that the accuracy is not high.Based on the problem of low accuracy of single-camera ball picking,a dual-camera ball picking algorithm is given,and an auxiliary camera is installed on the robot.The host computer uses the main camera to obtain the robot and the tennis.The robot rotates according to angle value.After the rotation,host computer reads auxiliary camera image data,analyze and calculate image data,calculate the angle value between current robot and the tennis ball to be picked up.According the angle value,it is judged whether the robot directly advances to pick up the tennis ball or picks up the tennis ball after further rotation,then,improve the accuracy of ball picking.(7)The experimental test of the intelligent ball picking robot is carried out;the path planning algorithm and the tennis picking algorithm are tested in combination.The combined algorithms are: Fusion of path planning algorithm and single camera ball picking algorithm(FS),Fusion path planning algorithm and dual camera ball picking algorithm(FT),Improved ant colony algorithm and single camera ball picking algorithm(AS),Improved ant colony algorithm and dual camera ball picking algorithm(AT).Among the four combined algorithms,the improved ant colony algorithm and single camera ball picking algorithm(AS)yields the shortest ball picking time while the improved ant colony algorithm and dual camera ball picking algorithm(AT)yields the highest probability of picking the ball.In comprehensive consideration,the improved ant colony algorithm and dual camera ball picking algorithm(AT)is adopted for this project.