Advances in the Validation and Application of Markerless Motion Capture Technology in Sports Performance and Clinical Rehabilitation

Technological Transformation: From Marker-Based Tracking to Computer Vision

Although traditional marker-based optical motion-capture systems remain the “gold standard,” their high equipment costs, cumbersome workflows, and reliance on laboratory settings have limited their adoption in clinical settings and on‑field sports applications. In recent years, markerless motion-capture systems powered by deep learning—such as OpenCap and Theia3D—have emerged rapidly, enabling three-dimensional kinematic analysis with nothing more than a smartphone and thereby reshaping the paradigms of athletic performance assessment and clinical rehabilitation. ^[1] 。

OpenCap System Validation: Geometric Accuracy?

Çabuk et al. in Biology of Sport Inclusion of a three-level meta-analysis published in (2026) 12 studies, 184 participants , extract 640 effect sizes . The correlation between OpenCap and the gold standard reaches r = 0.845 , the adjusted RMSE of the ensemble is 4.940° ^[2] 。

Cycling exercise verification

Kakavand et al. applied OpenCap to cycling analysis, achieving correlation coefficients of nearly 1 for the sagittal-plane angles of the hip, knee, and ankle. r > 0.9 , with RMSE values of less than 7.5°, 9.5°, and 11°, respectively. ^[3] , indicating that the markerless system has achieved high accuracy in periodic motion.

Challenges of Complex Upper-Extremity Movements

Thomas et al. (2025) compared the performance of Theia3D and the Qualisys system during throwing motions: the knee joint RMSD was only 7.17° ± 3.88° , but the wrist reaches as high as 26.66° (Old version). The new version reduces the elbow joint RMSD from 22.22° to 16.68° , the wrist drops to 18.05° ^[4] , algorithmic iteration is steadily narrowing the gap.

A New Data-Driven Paradigm for Motion Assessment

From High-Dimensional Data to a Comprehensive Score

Archibeck et al. (2025) introduced the Kinematic Synthesis Score (K-Score), which condenses whole-body motion data into a single metric. In a cohort of patients with chronic low back pain, the K-Score for the healthy control group was 94.16 ± 2.64 , the patient group is 85.82 ± 7.73 (p < 0.001), and is not confounded by BMI or age. ^[5] . This type of algorithm combines BOB Human Movement Biomechanics Analysis Software These platforms help standardize clinical assessments of motor function.

Computational Design Optimization of Sports Equipment

Kuzmeski et al. (2026) used motion-capture data to drive running-shoe design, and the prototype shoes were tested on 15 runners. Running economy improved by 3.1%–3.6% , outperforming top-tier racing shoes such as the Nike Alphafly 3, and all 15 athletes achieved their personal best performances. ^[6] 。

Open datasets accelerate clinical translation.

Baseline Data for Healthy Individuals

Vielemeyer et al. (2026) published 13 healthy adults A full-body motion capture dataset collected on slopes of 0°, 7.5°, and 10°, using a Vicon 10-camera system in conjunction with three force plates. ^[7] , providing an important reference for the development of assistive devices such as prosthetics and exoskeletons.

Large-scale postoperative rehabilitation data

Lunn et al. (2026) disclosed 137 patients who underwent total hip arthroplasty Motion capture and ground reaction force data during the execution of daily activities such as walking, squats, and stair climbing. ^[8] . The research team used AnyBody human musculoskeletal simulation modeling software A personalized musculoskeletal model is constructed and scaled based on anthropometric data, demonstrating a complete clinical workflow from motion-capture data acquisition to musculoskeletal simulation.

Trends and Prospects

Markerless motion capture technology is evolving along three main axes: Accuracy continues to improve. — Algorithmic iteration leads to a reduction in systematic error. ^[2][4] ； The scenarios are constantly expanding. — Extending from gait to cycling, throwing, and everyday activities ^[3][4][8] ； Mature data ecosystem — Open datasets accelerate cross-institutional collaboration ^[7][8] However, the meta-analysis also indicates that error variability remains substantial across different joints and tasks, and clinical validation data are still insufficient. ^[2] As deep-learning optimization and multimodal sensor fusion advance, markerless motion capture is poised to become a “ready-to-use” standard tool for motion analysis and rehabilitation assessment. ^[1] 。