A Computer Vision and Physics-Based Simulation Framework for Biomechanical Analysis and Predictive Modeling of Long Jump Performance
Description
This dataset and computational framework provide a comprehensive, non-invasive biomechanical analysis and predictive analytics system for long jump athletes. By integrating state-of-the-art computer vision (MediaPipe Pose) with rigid-body mechanics, mathematical filtering, and numerical aerodynamic models, this system establishes a reliable "Digital Twin" pipeline to evaluate approach, takeoff, flight, and landing phases. The implementation has been rigorously designed to ensure frame-by-frame data consistency, utilizing Kinovea standards as a validation baseline. Methodological Architecture: Pose Estimation & Landmark Trajectories: Captures sagittal-plane video data via generic camera units, mapping 33 distinct human skeletal landmarks across consecutive frames to produce structured 2D coordinate matrices. Noise Mitigation: Implements a Savitzky-Golay smoothing filter to eliminate digital tracking noise from the raw frame trajectories, crucial for steady velocity calculus. Anthropometric Calibration: Employs an enhanced multi-point scale calibration using the shoulder-to-ankle ratio against total athlete height, correcting pixel-to-meter coordinate transformations. Segmental Center of Mass (COM): Utilizes a segmental weighted method leveraging mathematically validated body-mass proportions (Dempster’s segment weights) to dynamically calculate the whole-body COM across the jump execution. Kinematic Variables & Takeoff Diagnostics: Automatically determines joint kinematics (angles for hips, knees, and ankles) and identifies the discrete takeoff moment through combined metrics: peak vertical velocity, abrupt vertical acceleration shifts, and foot ground-contact termination. Aerodynamic Flight Simulation: Models the airborne phase using projectile differential equations integrated with aerodynamic lift, drag forces, and local horizontal wind vector modifications via numerical Euler integration. Potential Reuse and Scientific Applications:This research data and code asset serve sports biomechanists, athletic coaches, and computer vision researchers. The dataset can be used to model sports performance optimizations, evaluate aerodynamic influences on human flight trajectories, or serve as a foundational testing dataset for human motion tracking models under rapid accelerations. Core Academic Foundations Reflected:Hay, J. G. (1993). The Biomechanics of Sports Techniques. Lees, A. et al. (1994). A Biomechanical Analysis of Touchdown and Takeoff Characteristics of the Men's Long Jump. Dempster, W. T. (1955). Space Requirements of the Seated Operator. Linthorne, N. P. (2006). Effect of Wind on the Long Jump.
Files
Institutions
- Sabaragamuwa University of Sri LankaSabaragamuwa Province, Ratnapura