3D Ball Trajectory from a Single Camera: A Cost-Effective Ball Estimation Technique for Cricket

Abstract

Accurate Three-Dimensional (3D) tracking of a cricket ball is a fundamental requirement for Decision Review Systems (DRS), yet existing solutions rely on expensive multi-camera infrastructures that limit accessibility, particularly in resource-constrained settings. This paper presents a cost-effective framework for estimating the 3D trajectory of a cricket ball using a single calibrated camera. The proposed method combines monocular vision geometry with a physics-based motion model to reconstruct ball position in world coordinates. Initial depth recovery is obtained from image-plane measurements and camera calibration, after which a Kalman filter incorporating a ballistic motion model is employed to suppress measurement noise and enforce physically plausible trajectories. The filtered trajectory enables reliable detection of the ball bounce point and forward prediction of post-bounce motion, which are critical for Leg-Before Wicket (LBW) analysis. Experimental results on real cricket delivery sequences demonstrate that the proposed Kalman-based smoothing reduces high-frequency reconstruction jitter by approximately 55%, decreasing the root mean square deviation from 42.7 mm to 19.2 mm. Despite using only a single camera, the reconstructed trajectories exhibit stable depth estimation and consistent bounce localization, highlighting the feasibility of the approach for low-cost DRS applications. The proposed system offers a practical alternative to multi-camera solutions, making 3D ball tracking more accessible for domestic cricket, training analysis and emerging cricketing regions.