Automated Measurement of Standing Long Jump Performance Using Binocular Stereo Vision

To overcome the inherent limitations of monocular vision systems in standing long jump measurement—including vulnerability to environmental noise and strict site-specific calibration requirements—as well as the implementation barriers of multi-modal sensor solutions characterized by excessive hardware costs and operational complexity, this paper presents an automated performance assessment framework based on binocular stereo vision with spatiotemporal trajectory modeling. The proposed system achieves robust environmental adaptability through two mechanisms: 1) depth-driven region-of-interest (ROI) filtering to exclude extraneous background elements beyond standardized jump areas, and 2) temporally consistent 3D coordinate tracking that directly reconstructs athletes' movement trajectories in physical space coordinates (x, y, z), eliminating manual pixel-to-physical calibration. This architecture enables precise landing point localization (mean error: 1.6 cm) across dynamically changing testing environments with heterogeneous illumination and background patterns. Experimental validation demonstrates significant advancements over conventional approaches, exhibiting 66.7% higher accuracy than monocular baselines while maintaining deployment flexibility unattainable by sensor-fusion alternatives. The integration of disparity-optimized depth perception and kinematic sequence analysis establishes a new paradigm for vision-based sports motion metrology.