A Pass Prediction System Derived from the Broadcasting Soccer Video

Date of Submission

September 2022

Date of Award

9-1-2023

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science

Department

Electronics and Communication Sciences Unit (ECSU-Kolkata)

Supervisor

Mukherjee, Dipti Prasad (ECSU-Kolkata; ISI)

Abstract (Summary of the Work)

In soccer, the most frequent event that occurs is a pass. For a trained eye, there are a myriad of adjectives which could describe the possible pass (e.g.,"majestic pass", "conservative" to "poor-ball"). During the game, defending players constantly try to predict the pass of the attacking player to prevent a goal. So, pass prediction is an important facet to anticipate the game strategy of the participating teams. Related work in this area relies on the information extracted from multiple camera installations in the stadium or data coded by human annotators sitting in the stadium. Aberrating the state-of-the-art, we present a pass prediction system directly derived from the broadcasting video. In order to develop the pass prediction system, three components are needed as follows. The first component is to track the ball in a given broadcasting video. The problem of tracking ball in a soccer video is challenging because of sudden change in speed and orientation of the soccer ball. Successful tracking in such a scenario depends on the ability of the algorithm to balance prior constraints continuously against the evidence garnered from the sequences of images. In this thesis we propose a particle filter based algorithm that tracks the ball when it changes its direction suddenly or takes high speed. Exact, deterministic tracking algorithms based on discretized functional, suffer from severe limitations in the form of prior constraints. In contrast, our tracking algorithm exploits a probabilistic framework and has shown excellent result even for partial occlusion. A holy grail for sports analytics is the top-view visualization of the game. The top-view visualization provides the actual between-player distances as opposed to the between-player distances calculated from the side and/or oblique view of a match as shown in the broadcasting video. Therefore the second component of the pass prediction system is the top-view visualization of the match. In this work, we present a factor theory based approach to derive the top-view visualization of the game from the broadcasting sports video. We theoretically prove that the proposed factor theory based approach for top-view visualization is more efficient than the state-of-the-art approach. In addition, as per the proposed approach, we present a model for the top-view visualization by transforming the broadcasting video into a single and static camera visualization. In order to generate the single-camera visualization, the view of the entire ground is needed which is expressed as a solution to a convex optimization function, devised to explore putative matrix completions. Finally we present a probabilistic framework for pass prediction. The proposed framework predicts pass recipients by integrating two dependent models, designed from the coordinates of the players in abstract top-view visualization. The contribution of the work are generation of the proximity model based on the positions of the opponent team players, and generation of pass region model that is influenced by the concentration of the players of the team who is in possession of the ball. To evaluate the real time efficacy of the proposed pass prediction system, a soccer data set has been introduced. The proposed pass prediction system is compared against recent methods and the ground truth available in the soccer data set. The proposed method outperforms the existing approaches by a noticeable margin

Comments

ProQuest Collection ID: https://www.proquest.com/pqdtlocal1010185/dissertations/fromDatabasesLayer?accountid=27563

Control Number

ISILib-TH558

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

DOI

http://dspace.isical.ac.in:8080/jspui/handle/10263/2146

Link to Full Text

Share

COinS