What is it about?

In sports, accurately mapping 2D images from cameras to real 3D positions on the field is essential. This precision enables tasks such as player tracking, offside detection, performance analysis, and more. Our AI team at Sportlight.ai has developed a method to achieve this by closely analyzing the lines and curves on the soccer field, yielding excellent results. Using AI models, we identify critical field markings and align them with known features of real soccer fields. This approach earned us first place in the SoccerNet Camera Calibration Challenge 2023, which remains the largest available soccer broadcast camera calibration dataset.

Featured Image

Why is it important?

Our method addresses the critical challenge of finding a sufficient number of high-quality point pairs for accurate calibration by leveraging the inherent structural features of the football pitch. This includes exploiting line-line and line-conic intersections, points on the conics themselves, and other geometric features of a football pitch, thereby significantly increasing the number of usable points and enhancing both accuracy and robustness.

Perspectives

We believe this research shows that you don't always need complex methods to get great results. By carefully choosing a small set of important points on the field and using straightforward computer analysis, we were able to outperform more complicated approaches. Our method works well for soccer, but we think it could be useful for other sports too. Any sport that uses lines or curves on its playing field could potentially benefit from our ideas.

Ruilong Chen

Read the Original

This page is a summary of: Enhancing Soccer Camera Calibration Through Keypoint Exploitation, October 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3689061.3689074.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page