Team-Aware Football Player Tracking with SAM: An Appearance-Based Approach to Occlusion Recovery

TL;DR

The paper tackles football player tracking under frequent occlusions and uniform similarity by proposing a lightweight pipeline that combines SAM-based initialization with CSRT tracking and jersey-color appearance for re-identification. It presents a team-aware approach that uses domain-specific cues to enhance occlusion recovery while maintaining computational efficiency suitable for post-match analysis. A multi-dimensional evaluation framework assesses speed, accuracy, and robustness, revealing strong performance in light/moderate occlusions but limited long-term re-acquisition, highlighting the need for memory-based re-identification. Practical guidelines are provided for deploying such systems under resource constraints, and the work points to future extensions with memory-enabled SAM variants and enhanced re-identification strategies for improved long-duration robustness.

Abstract

Football player tracking is challenged by frequent occlusions, similar appearances, and rapid motion in crowded scenes. This paper presents a lightweight SAM-based tracking method combining the Segment Anything Model (SAM) with CSRT trackers and jersey color-based appearance models. We propose a team-aware tracking system that uses SAM for precise initialization and HSV histogram-based re-identification to improve occlusion recovery. Our evaluation measures three dimensions: processing speed (FPS and memory), tracking accuracy (success rate and box stability), and robustness (occlusion recovery and identity consistency). Experiments on football video sequences show that the approach achieves 7.6-7.7 FPS with stable memory usage (~1880 MB), maintaining 100 percent tracking success in light occlusions and 90 percent in crowded penalty-box scenarios with 5 or more players. Appearance-based re-identification recovers 50 percent of heavy occlusions, demonstrating the value of domain-specific cues. Analysis reveals key trade-offs: the SAM + CSRT combination provides consistent performance across crowd densities but struggles with long-term occlusions where players leave the frame, achieving only 8.66 percent re-acquisition success. These results offer practical guidelines for deploying football tracking systems under resource constraints, showing that classical tracker-based methods work well with continuous visibility but require stronger re-identification mechanisms for extended absences.

Figures (5)

• Figure 1: Players to be tracked can be selected upon user click, with the ability of selecting between 1) Team 1 2) Team 2 3) Referee (Assistant). The point prompts are passed to the model to segment and proceed with tracking
• Figure 2: Tracked players are marked with a bounding along with their movement path. Labels of the team are included as well
• Figure 3: Light occlusion scenario showing typical gameplay with 1-2 players in close proximity. Tracked players are marked with bounding boxes and team labels.
• Figure 4: Heavy occlusion scenario in the penalty box with 5+ players in close proximity. Multiple overlapping players create challenging tracking conditions with frequent identity ambiguity.
• Figure 5: Long-term occlusion scenario where a player (goal keeper) exit and re-enter the camera view. The system must maintain player identity across extended absences from the frame.