Automatic detection of ball and players in soccer games

• Main Author: Yu, Kangran.
• Other Authors: Foo Say Wei
• Format: Final Year Project
• Language: English
• Published: 2011
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: http://hdl.handle.net/10356/46030
• Institution: Nanyang Technological University

Automatic analysis of ball and player position from a match video is an essential image and video processing topic. Moreover, it also plays a significant role in professional coaching and training of soccer nowadays. This project aims to study the success and failure of past researchers and develop new methodologies that perform ball and players’ positions detection from a video file. The detection is discussed under two occasions, when target is not overlapped by other objects and when occlusion takes place. Regarding these scenarios, two different Matlab based algorithms are established to accomplish detection/estimation of ball and players position. One of the methods is based on the intuitive image processing and a new roundness computation while the other algorithm executes candidate segments comparison by estimating covariance matrices and eigenvectors. The basic theory of multivariate distribution in statistics and probabilities is also implemented in the latter algorithm. After all, two algorithms are constructed into one function to perform active and reliable detection in both occlusion and non-overlapped cases. Additionally, this project only focuses on detection of objects on a 2-dimensional space, thus match videos of Barcelona F.C., Arsenal F.C., and Real Madrid are selected as samples thanks to their relatively fluid and ground based match flow.