Automatic soccer field of play registration

摘要

This paper proposes a strategy for the automatic registration of soccer images on a model of the field of play. First, a robust and efficient preprocessing is applied to discard the areas of the image that do not belong to the field of play and eliminate most edge points that are not part of the line marks. Then, a novel probabilistic decision tree is used to identify the most probable classification for the set of all the straight lines in the image. Additionally, the line surrounding the center circle is also modeled for providing results when only few straight lines are visible. Finally, a three-step analysis stage is applied to ensure the validity of the results.