Automatic Conversion of Broadcasted Football Match Recordings to Its 2D Top View

This paper presents an automated method to convert a broadcasted football match recording to its 2D Birds eye view. Representing a football match in 2D can be useful for tactical analysis and generating stats from events in the football match. We perform player detection and tracking on the broadcasted recording using YOLOv7 and DeepSort tracker. A database generation process is discussed to generate images of different views and perspectives of the football field along with their perspective transformation matrices. After performing some pre-processing steps on the frames from the broadcasted recording, it is compared with the images in the database to find the closest match. The perspective transformation matrix of the closest match is then used to convert the input image to its 2D top view.