Estimation of position and trajectory of a flying ball in 3-D space

This paper proposes an algorithm for estimating the position and trajectory of a flying ball in 3-D space with a RGBD (RGB and Depth) sensor. An integrated ball and its trajectory detection using color-based back projection method is proposed. The k-means clustering is also employed to make the detector more robust even with noise in the background. The integrated detector does not need any type of prior knowledge about the template of the object, and still has a good performance in different lighting conditions, and with low quality images. We introduce a confidence parameter for the work area of the manipulator where it must move to catch the ball. The experimental results suggest that the integrated detector can predict the trajectory of the ball, and also estimate the position in 3-D space with high accuracy.