A Feasibility Study of Video-Based Performance Data Extraction for Football Referee Self-Analysis Applications

Sammanfattning

Football is one of the world’s most popular sports, yet the development of performance analysistools has primarily focused on players and teams, leaving limited support for match officials. Forfootball referees, positioning and proximity to active play are important factors for decision-making.This study addresses this gap through three connected research questions concerning domainunderstanding, technical feasibility, and the design and evaluation of a referee self-analysis conceptbased on objective spatial performance metrics from standardized match recordings. The study combines key performance indicator (KPI) identification, technical development,conceptual application design, and expert evaluation to assess both the technical viability of theproposed system and its potential practical value for referee development. These parts correspondto the three research questions by first identifying relevant referee performance indicators, thenassessing whether objective metrics can be extracted from match recordings, and finally exploringhow the extracted data can be presented and evaluated in a self-analysis prototype. Based onliterature, domain knowledge, and expert input, relevant referee performance indicators wereidentified and used to guide the development of the Proof of Concept. By implementing computervision techniques and homography calibration, the system tracks the referee’s movement andprojects video coordinates onto a two-dimensional representation of the pitch. In addition to thetechnical pipeline, the study presents a conceptual design for a web-based self-analysis applicationtailored to referees. The results indicate that it is technically feasible to generate positioning heatmaps and calculatebasic movement metrics using existing camera infrastructure. In the technical analysis, 17 videoclips from two different matches were processed. For the primary dataset, consisting of 14 clips, thesystem calculated a total covered distance of 1,027.7 meters. An isolated five-second sprint intervalwas also analyzed, in which the system calculated a covered distance of 23.3 meters. Furthermore,the expert evaluation suggests that combining objective spatial data with a user-centeredapplication design has potential to support referee education and self-analysis. The study concludesthat accessible video-based spatial analysis can provide a foundation for future tools aimed atimproving referee development outside elite-level contexts.