A Monocular Ball Localization Method for Automatic Hitting Robots

The batting robot has to meet the requirements of high-speed continuous motion as well as quickly recognize and localize the table tennis ball and predict the landing point in the working process. The recognition accuracy of the ball in this process is an important factor in measuring the performance of the batting robot. In order to accurately improve the ball localization accuracy, a ball recognition and localization method based on the elliptic projection transform and test field joint correction fitting is proposed. The method uses two monocular cameras of the same model to collect the sphere motion trajectory data at low cost and improves the localization accuracy of the table tennis ball during its motion by combining the elliptic projection transform of the sphere on the phase plane of the monocular camera with the test field data. On this basis, the obtained height information is fitted to the ball motion trajectory to realize the prediction of the table tennis ball's hitting point.