A cutting-edge system for managing a professional team in real-time, akin to a virtual coach

Аннотация

Objective of the study was to creating a dynamic game management system that operates in real-time for a professional team («Computer Coach»).

Methods and structure of the study. In the experimental phase of the research, a comparative evaluation of the efficiency of computer vision algorithms employed to analyze video footage of basketball games in the ASB Division championship in the Sverdlovsk region was conducted. The data obtained were used to assess the precision and performance of mathematical models designed to address the challenges of the control system.

Results and conclusions. The objectives of calculating the three-dimensional coordinates of players with an accuracy of 0.4 meters, identifying players, the ball, tracking players, categorizing teams of players, and recognizing numbers on jerseys were achieved. Without the use of neural networks, it was possible to identify 30 types of technical and tactical martial arts (TTA), a method for generating voice commands in real-time, and elements of communication with players were implemented. We estimate the potential impact of the implementation at 22-28% of the additional points scored, however, due to a number of technical constraints, it is not yet possible to fully utilize the project's potential. The authors are eager to collaborate with professional clubs to complete the project.