Â鶹ԼÅÄ

Analysis of sports such as football often makes use of techniques to allow the broadcaster to show a view of incidents in a match from different angles. This can involve a manual process of creating a model of the game at a key moment, requiring a 3D model of each player to be created, or selected from a pre-defined library of players in different poses. This is time-consuming, limiting the number of incidents that can be analysed.

This paper presents a method to automatically identify the pose and orientation of a football player from a video still image to allow such applications to be speeded up.

The Random Fern Forest classifier is used in conjunction with a set of 3D person models to produce a pose estimation of the football player.

This is a quick classifier so can be used in applications that are required to run in a very short time. This document was originally presented and published at the Intelligent Signal Processing (ISP) Conference, London 1st & 2nd December 2015. The slides are included in an appendix to the white paper.

-

Â鶹ԼÅÄ R&D - Intelligent Video Production Tools

Â鶹ԼÅÄ R&D - Piero

Â鶹ԼÅÄ R&D - The Queen's Award winning Piero sports graphics system

Â鶹ԼÅÄ Internet Blog - Â鶹ԼÅÄ R&D wins Queen's Award for Enterprise for Piero

Â鶹ԼÅÄ R&D - Olympic Diving 'Splashometer'

Â鶹ԼÅÄ R&D - Biomechanics

Â鶹ԼÅÄ R&D - Real-time camera tracking using sports pitch markings

Â鶹ԼÅÄ R&D - Augmented Reality Athletics

Â鶹ԼÅÄ R&D - Image-based camera tracking for Athletics

Â鶹ԼÅÄ R&D - Use of 3-D techniques for virtual production