Robust Automatic Camera Calibration in Badminton Court Recognition

Badminton has attracted a large number of enthusiasts with its unlimited venues and inexpensive nature, and badminton-related studies have emerged, all of which mainly focus on the analysis of badminton videos. Badminton video analysis requires the foundation of court recognition, and this paper discusses the recognition of badminton courts in images. We propose an automatic camera calibration algorithm that is robust in badminton court recognition, which can recognize badminton courts under a variety of shooting angles and solve the problem of court line mapping in the case that the real court lines fall behind the camera optical center. The main process of our algorithm includes three parts: candidate pixel extraction, court line detection and model fitting, adding line number constraints and 3-dimensional mapping filtering to the traditional camera calibration model to improve the robustness of the algorithm. Our algorithm is tested on a self-built dataset containing 300 images of badminton courts from different angles and has a good performance. Our algorithm can be applied to court recognition of other ball sports and can also be extended to court recognition in videos by modifying the standard model, with a wide range of application scenarios.