Study Of Autonomous Mobile And Target Recognition And Tracking Of Basketball Robot

With the development of science and technology,robots have been widely used in industry,service industry and entertainment industry.Basketball robots are an important project in robot competitions and involve many hot topics in robot research.This paper uses the basketball robot as the experimental platform,and mainly studies four issues: the hardware platform construction of the basketball robot,the robot selflocalization algorithm,the target recognition algorithm and the target tracking algorithm.Firstly,a robot hardware platform is built based on the rules of the basketball robot competition and the kinematics analysis of the robot chassis is performed.The basketball robot is divided into functional modules,the four-wheel mecanum wheel O-rectangular structure is used to build the chassis,and the kinematics analysis of the robot chassis is carried out.Secondly,the self-localization algorithm of the basketball robot is studied and the experimental test is completed.The Kalman filter algorithm and the extended Kalman filter algorithm are studied,the system model of the odometer and IMU is established,and the data fusion of the odometer and IMU is carried out using the extended Kalman filter algorithm which is more suitable for nonlinear systems.Experiments verify that the position data of the basketball robot is more accurate after using the self-localization algorithm.Then the target recognition algorithm of the basketball robot is studied and the experimental test is completed.Research on the recognition algorithm based on color and contour,and experimental tests is conducted to point out its shortcomings.Then the SSD(Single Shot MultiBox Detector)algorithm is studied.The MobileNet network is used to replace the backbone network in the SSD,which reduces the amount of model calculations and the number of parameters;the priori box setting in the SSD algorithm is changed to better identify the round ball.Experimental tests based on the color and contour algorithm and SSD-MobileNet algorithm prove that the latter is more accurate than the former in identifying the target ball under strong light,low light and complex background.Finally,the target tracking algorithm of the basketball robot is studied and the experimental test is completed.For the Kernelized correlation filter tracking algorithm,the maximum response value index and the Average Peak-to-Correlation Energy are proposed to determine whether the tracking target is lost.If the tracking is lost,the SSD-MobileNet algorithm is used to re-identify the target,and the KCF algorithm model is updated with the information of the target ball.Experiments verify that the algorithm can effectively complete the tracking when the target ball translation,scale transformation and occlusion.The basketball robot designed based on the research method in this article has been tested in actual combat in the China Robot Competition and achieved good results,which also verifies the effectiveness of the research method in this article.