EUVIMEDEUVIMED entdecken
Crossref

Algorithm for Identifying Athlete Passing Trajectories in Team Competition Scenarios

Frontiers in Sport Research · 2026 · 8 (1)

DOI: 10.25236/fsr.2026.080109 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Sport Research, 2026.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

: Existing athlete passing trajectory recognition algorithms primarily rely on traditional object detection and single-target tracking frameworks, and these algorithms match trajectories and determining passing events based on kinematic features. However, these approaches only perform local optimization for individual athlete trajectories and lack global modeling of spatio-temporal correlations among multiple trajectories, resulting in low recognition accuracy. To address this, we propose an athlete passing trajectory recognition algorithm for group-based competitive scenarios. The algorithm constructs composite model of multiple Gaussian distributions is constructed for each pixel to represent the temporal multi-state characteristics of background pixels. A weight-sorting mechanism is designed to selects Gaussian model combinations with cumulative probability thresholds to form the background model, effectively separating athletes from background regions. Then the algorithm adopts a Kalman filter algorithm constructs an athlete state space model incorporating position and velocity, utilizing a state transition matrix for temporal trajectory prediction. Process noise covariance adapts to the rapid directional changes characteristic of athletes in group confrontations, enabling pass trajectory prediction. Passing events are detected by establishing a dual-condition constraint: when two athlete trajectories satisfy spatial proximity and directional consistency within a spatiotemporal neighborhood, they are classified as potential passing events. By backwardly reconstructing the passer's trajectory segment and forwardly tracing the receiver's trajectory segment, the complete passing path is formed through temporal stitching. Finally, the algorithm applies nonlinear optimization globally corrects the merged trajectories by minimizing the sum of squared residuals between observed and predicted states, enabling pass trajectory recognition. Experimental validation confirms the proposed method's recognition accuracy. Comparative test results demonstrate that when applied to athlete pass trajectory recognition, the proposed approach achieves 92.5% match accuracy between predicted and actual trajectories, delivering highly satisfactory recognition performance

Zitieren

APA

(2026). Algorithm for Identifying Athlete Passing Trajectories in Team Competition Scenarios. Frontiers in Sport Research, 8(1). https://doi.org/10.25236/fsr.2026.080109

Vancouver

Algorithm for Identifying Athlete Passing Trajectories in Team Competition Scenarios. Frontiers in Sport Research. 2026;8(1). doi:10.25236/fsr.2026.080109

BibTeX
@article{euvimed2026,
  title = {Algorithm for Identifying Athlete Passing Trajectories in Team Competition Scenarios},
  journal = {Frontiers in Sport Research},
  year = {2026},
  volume = {8},
  number = {1},
  doi = {10.25236/fsr.2026.080109}
}
Kostenlos starten →