Fundamental Research on Athlete Positions Estimation in Indoor Sports at Various View

In recent years, data collection for tactical analysis in sports has become increasingly prevalent.In sports such as volleyball, basketball, and soccer, where player positioning is closely linked to scoring opportunities, research has been conducted to visualize player positions using various technological approaches.Notably, numerous research has focused on enhancing tactical analysis by estimating player positions through image recognition methodologies.These approaches typically rely on images captured by one or more cameras.From these images, specific reference points on the court are identified and transformed into a bird's-eye view using image transformation algorithms such as projective transformation to visualize player positioning.This process requires the selection of four reference points on the court, preferably encompassing the entire playing area.However, capturing these four reference points from ideal viewing angles is often infeasible in many venues.Additionally, live game images frequently feature zoomedin views of players and shifts in camera angles as the ball is tracked, creating challenges for consistent analysis.These limitations restrict tactical analysis to videos specifically recorded for such purposes, excluding archival video not originally intended for research and thus limiting data diversity.To address these challenges, this research aims to develop a system that broadens the applicability of tactical analysis by utilizing video captured in stadiums where a part of views is feasible, as well as historical video data.Focusing on volleyball, the proposed approach automatically identifies reference points based on the coordinates of the net and court lines to estimate player positions.This system seeks to enable robust and efficient analysis across diverse video sources, enhancing the scope and utility of tactical insights.