Kalman tracking algorithm based on real-time vision of ping-pong robot

An adapted measurement covariance digital Kalman filter method was proposed to eliminate the locating noises caused by motion blurring, air-drag and camera image distortion. The method implemented a precise tracking of the target motion trajectory by dynamically changing the measurement covariance, and laid a foundation for target prediction and arm motion. Experimental data showed a good tracking result in condition of camera capturing rate over 70 frame/s and ball motion speed over 5 m/s since the method effectively suppressed interference of noisy measurement and data losing. The method can be employed in cases of fast object tracking for its low computation load and high tracking precision.