Field-Based Evaluation of a Three-Camera Video-Based System for Three-Dimensional Kinematic Analysis of Powerlifting Movements
Abstract
Objective: Three-dimensional (3D) kinematic assessment is important for quantifying movement mechanics during resistance exercise; however, laboratory-based motion-capture systems are often costly, technically demanding, and difficult to implement in field settings. Low-cost multi-camera video-based systems may offer a practical alternative, but their measurement agreement during complex multi-joint resistance exercises remains insufficiently established. This study aimed to evaluate the agreement between a three-camera video-based 3D kinematic analysis system and an expert manual video-digitization reference method during squat, bench press, and deadlift exercises performed under field-based conditions.
Methods: This cross-sectional agreement study included 42 male university powerlifters who performed competition-standard squat, bench press, and deadlift trials. Movements were recorded using three synchronized digital video cameras operating at 120 Hz and analyzed with Kinovea software. The proposed system was compared with an expert manual video-digitization reference method performed by blinded analysts. Selected joint angles and mean concentric barbell velocity were analyzed. Agreement between methods was assessed using Pearson correlation coefficients, intraclass correlation coefficients (ICC), root mean square error (RMSE), paired-samples t-tests, and Bland–Altman analysis.
Results: Strong correlations were observed across all measured variables (r = 0.946–0.982, p < 0.001), and agreement was excellent based on ICC values ranging from 0.944 to 0.981. RMSE values were low, ranging from 1.63° to 1.91° for joint-angle measures and from 0.015 to 0.016 m·s⁻¹ for barbell velocity. Paired-samples comparisons showed no statistically significant differences between the proposed system and the expert manual reference method for seven of the eight variables. Although the bench press elbow angle showed a statistically significant difference, the magnitude of this difference was less than 1° and was not considered practically meaningful. Bland–Altman analysis indicated minimal systematic bias and narrow limits of agreement across the measured variables.
Conclusion: The proposed three-camera video-based system demonstrated high agreement with an expert manual video-digitization reference method for selected 3D kinematic variables during squat, bench press, and deadlift exercises under standardized field-based conditions. Although these findings should not be interpreted as validation against laboratory-grade optical motion-capture systems, the system may provide a practical option for field-based biomechanical assessment when laboratory motion-capture facilities are unavailable.
Introduction
Powerlifting is a maximal strength sport consisting of three movements—squat, bench press, and deadlift—that rely on the athlete's ability to generate high levels of force while maintaining technically correct movement patterns. Beyond maximal strength, biomechanics is a key determinant of performance because it affects movement efficiency, load distribution, and injury risk. Barbell leverage, joint kinematics, and segmental coordination can substantially influence lift success, especially under near-maximal loading conditions [1].
Three-dimensional (3D) kinematic analysis is commonly used in biomechanics to describe motion patterns and analyze the mechanical characteristics of sport-specific tasks. Accurate measurement of joint angles, segmental motion, and barbell displacement is important for technique analysis and optimization [2,3]. Because of their high spatial and temporal resolution, laboratory-based motion-capture systems such as Vicon and Qualisys are often considered reference standards [4]. However, their cost, technical complexity, and requirement for controlled laboratory conditions limit their practical use in routine training, competition settings, and large-scale applied research [5].
In recent years, other methods to assess motion have emerged that do not require reflective markers or specific laboratory facilities, providing an alternative method. Multi-camera video systems and motion analysis software have shown to be a viable approach for a variety of activities in assessing human motion [6–8]. These can include more convenient access and facilitate data collection in environments that are more natural to the sport, such as resistance training and strength sports.
Despite these advances, the accuracy of low-cost video-based systems for complex multi-joint resistance exercises remains insufficiently established. Powerlifting movements create specific methodological challenges, including large ranges of motion, rapid changes in joint position, and partial occlusion of anatomical landmarks by the body, barbell, or external load [9,10]. Previous validation and agreement studies have largely focused on gait, running, or low-load activities, with limited evidence for high-load resistance exercises conducted in realistic field settings [11]. This creates a practical need for accessible and objective movement-analysis methods in strength-training environments [12].
This study addresses this gap by assessing the agreement of a low-cost multi-camera video-based system for 3D kinematic analysis of powerlifting movements. Although video-based motion analysis is not novel, the present investigation evaluates measurement agreement under ecologically valid resistance-training conditions involving multiple planes of movement and high external loads. By focusing on selected joint kinematics and barbell velocity and by applying multiple agreement metrics, the study should be interpreted as a field-based measurement agreement assessment rather than a technological validation against laboratory-grade optical motion capture.
This study contributes to the growing body of research on accessible three-dimensional (3D) motion analysis by evaluating a low-cost three-camera video-based system under field-based conditions. Unlike previous studies that have primarily focused on automated markerless motion capture or laboratory-based optical systems, the present investigation examined the agreement between a synchronized three-camera video-based approach and an expert manual video-digitization reference method for complex multi-joint powerlifting movements. The study focused on selected joint kinematics and barbell velocity and employed multiple agreement metrics, including intraclass correlation coefficients (ICC), root mean square error (RMSE), paired-samples comparisons, and Bland–Altman analysis, to comprehensively evaluate measurement agreement.
The aim of this study was to evaluate the agreement between a three-camera video-based system and an expert manual video-digitization reference method for three-dimensional (3D) kinematic analysis of squat, bench press, and deadlift exercises under standardized field-based conditions. Because laboratory-based optical motion capture systems were not available in the present field setting, expert manual video digitization was adopted as the practical reference method for comparison. Accordingly, the findings should be interpreted as evidence of agreement with this reference method rather than criterion validation against a laboratory-grade optical motion capture system. It was hypothesized that the proposed system would demonstrate high agreement with the expert manual reference method, low measurement error, and sufficient accuracy to support practical biomechanical assessment in field environments where laboratory motion-capture systems are not readily available.
Materials and Methods
Study Design
A cross-sectional agreement study was conducted to evaluate the agreement between the proposed three-camera video-based system and an expert manual video-digitization reference method for three-dimensional (3D) kinematic analysis of powerlifting movements. Data acquired from the proposed three-camera video-based system were compared with an expert manual video-digitization reference method.
Participants
Forty-two male university powerlifters volunteered to participate in the study (age: 24.0 ± 0.8 years, height: 171.8 ± 4.6 cm, body mass: 74.9 ± 1.5 kg, training experience: 4.0 ± 1.2 years). Participants competed in the 75-kg weight category and had experience in structured resistance training and competitive lifting. Inclusion criteria were age between 23 and 25 years, at least two years of resistance-training experience, regular performance of the squat, bench press, and deadlift during the previous six months, and no history of musculoskeletal injury. Participants were excluded if they reported current injury or pain that affected movement execution, had incomplete data, or were unable to complete all testing procedures.
Ethical Approval
The study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Human Research Ethics Committee (Approval No. HITS/DRC/22701008/2026). Written informed consent was obtained from all participants before participation.
Experimental Setting and Procedures
Data collection was conducted in an indoor strength-training facility under standardized conditions with adequate lighting. Participants performed the squat, bench press, and deadlift according to standardized competition rules. Before testing, all participants completed a standardized warm-up consisting of dynamic mobility exercises followed by progressive lift-specific warm-up sets. One trial was recorded for each participant during the squat, bench press, and deadlift exercises using a standardized load corresponding to 80% of the individual's one-repetition maximum (1RM). This submaximal intensity was selected to ensure safe movement execution while representing a typical training load for competitive powerlifters and allowing consistent assessment of joint kinematics and barbell velocity across participants. A single trial was analysed because the primary objective of the study was to evaluate measurement agreement between the proposed three-camera video-based system and the expert manual video-digitization reference method under standardized field-based conditions rather than to assess within-session variability or trial-to-trial repeatability. To minimize the influence of fatigue, participants were provided with 3–5 minutes of passive recovery between lifts.
Three-Camera Video-Based System
Three synchronized digital video cameras (120 Hz, 1080p resolution) were positioned on tripods around the lifting platform to capture each lifting trial from three different perspectives. Camera 1 was positioned in the sagittal plane (90°), Camera 2 in the frontal plane (0°), and Camera 3 in an oblique anterior-lateral position (45°). All cameras were placed approximately 4 m from the centre of the lifting platform at a height of 1.2 m above the floor. Synchronization between the three cameras was achieved using a hand-clap immediately before each lifting trial. The synchronized recordings provided multiple views for subsequent manual landmark digitization and three-dimensional reconstruction. The experimental setup, camera positions, and calibration frame are illustrated in Figure 1.
Calibration and Measurement Volume
Prior to data collection, a cubic calibration frame (2.0 m × 2.0 m × 2.0 m) with known dimensions was positioned at the centre of the lifting platform to define the measurement volume and establish a common spatial reference for all three camera views. The calibration frame was recorded before participant testing to ensure consistent camera geometry throughout data collection. This procedure enabled reconstruction of three-dimensional movement coordinates within the predefined measurement volume. Quantitative calibration error, reprojection error, and reconstruction accuracy were not computed because the primary objective of the study was to evaluate measurement agreement between the proposed three-camera video-based system and an expert manual video-digitization reference method under standardized field-based conditions.
Motion Analysis Procedures
Video recordings acquired from the three synchronized cameras were analysed using Kinovea (Version 2025.1.1). Anatomical landmarks and bilateral barbell endpoints were manually digitized frame by frame in each camera view using standardized digitization procedures [13–15]. The anatomical landmarks included the acromion process, lateral epicondyle of the humerus, wrist joint centre, greater trochanter, lateral femoral condyle, lateral malleolus, and bilateral barbell endpoints. Three-dimensional coordinates were reconstructed by triangulating corresponding two-dimensional landmark coordinates obtained from the synchronized camera views within the calibrated measurement volume. Joint angles and barbell kinematic variables were then calculated from the reconstructed three-dimensional coordinates and analyzed using SPSS v26.0. No lens-distortion correction was applied during video processing. The overall workflow of data acquisition, synchronization, calibration, manual landmark digitization, three-dimensional reconstruction, and statistical analysis is illustrated in Figure 2.
Figure 2. Workflow of data acquisition and motion analysis
Kinematic Variables
The following kinematic variables were analyzed: peak knee flexion angle (°), hip flexion angle (°), and mean concentric barbell velocity (m·s⁻¹) for the squat; elbow flexion angle (°) and mean concentric barbell velocity (m·s⁻¹) for the bench press; and knee angle at lift-off (°), hip angle at lift-off (°), and mean concentric barbell velocity (m·s⁻¹) for the deadlift.
Expert Manual Video-Digitization Reference Method
Reference measurements were independently obtained by two experienced biomechanical motion analysis analysts who were not involved in the data collection process. Both analysts were blinded to the measurements obtained from the proposed three-camera video-based system. The same anatomical landmarks and standardized frame-by-frame manual digitization procedures were applied by both analysts, and the final reference values for each variable were calculated as the average of the two independent analyses to minimize observer-related variability.
Laboratory-based optical motion capture systems are widely regarded as the criterion standard for three-dimensional biomechanical analysis because of their high spatial and temporal accuracy [6]. However, such systems were not available under the field-based conditions of the present investigation. Therefore, expert manual video digitization was adopted as the practical reference method for comparison. Previous studies have demonstrated that manual digitization performed by experienced analysts using standardized procedures can achieve high levels of reliability and measurement consistency [9]. Accordingly, the objective of the present study was to evaluate the agreement between the proposed three-camera video-based system and an expert manual video-digitization reference method under standardized field-based conditions rather than to establish criterion validity against a laboratory-grade optical motion capture system. The findings should therefore be interpreted as evidence of agreement with an expert manual reference method within the specific experimental conditions investigated.
Reliability Assessment
Intra-rater reliability was evaluated by repeating the ratings for 15 randomly selected trials after seven days by the same evaluator. Inter-rater reliability was assessed by comparing measurements obtained from the two independent analysts [17]. Reliability outcomes were used to support the interpretation of measurement agreement and to ensure that observed differences between methods were not primarily attributable to observer variability.
Statistical Analysis
Statistical analyses were performed using IBM SPSS Statistics (Version 26.0; IBM Corp., Armonk, NY, USA). Data normality was assessed using the Shapiro–Wilk test and Q–Q plots. Descriptive statistics are presented as mean ± standard deviation (SD). Pearson correlation coefficients (r) and two-way random-effects intraclass correlation coefficients (ICC2,1) with 95% confidence intervals were calculated to evaluate agreement between the proposed three-camera video-based system and the expert manual video-digitization reference method [18]. Paired-samples t-tests were used to assess systematic differences between methods, while root mean square error (RMSE) and Bland–Altman analysis (mean bias and 95% limits of agreement) were used to evaluate measurement error and agreement. Statistical significance was set at p < 0.05. ICC values were interpreted as poor (<0.50), moderate (0.50–0.75), good (0.75–0.90), and excellent (>0.90) according to established guidelines, whereas lower RMSE values, smaller mean bias, and narrower 95% limits of agreement indicated closer agreement between the two measurement methods.
Results
No participants (N = 42) had missing data for the testing protocol. All variables were approximately normally distributed according to the Shapiro–Wilk test and Q–Q plot inspection (p > 0.05). Participant characteristics and performance variables are presented in Table 1. Table 2 summarizes the agreement and reliability analyses for the proposed three-camera video-based system compared with the expert manual video-digitization reference method, whereas Table 3 presents the Bland–Altman agreement statistics.
Variable | Mean ± SD | Min | Max |
|---|---|---|---|
Age (years) | 24.0 ± 0.8 | 23 | 25 |
Height (cm) | 171.8 ± 4.6 | 163.9 | 181.2 |
Body Mass (kg) | 74.9 ± 1.5 | 72.1 | 77.8 |
Training Experience (years) | 4.0 ± 1.2 | 2.1 | 5.9 |
Squat (kg) | 180.7 ± 18.4 | 151.6 | 209.4 |
Bench Press (kg) | 119.6 ± 14.1 | 95.8 | 144.3 |
Deadlift (kg) | 215.8 ± 20.9 | 181.7 | 248.6 |
Total (kg) | 516.1 ± 34.7 | 450.8 | 584.2 |
Note. Values are presented as mean ± standard deviation (SD).
Participants in the 75-kg weight category demonstrated competitive strength levels across all three lifts (Table 1). The range of performance measures was consistent with trained powerlifters and supported the suitability of the sample for biomechanical analysis.
Variable | Proposed System Mean ± SD | Reference Mean ± SD | r | ICC (95% CI) | Intra-rater ICC (95% CI) | Inter-rater ICC (95% CI) | SEM | MDC95 | RMSE | p-value |
|---|---|---|---|---|---|---|---|---|---|---|
Squat Knee Angle (°) | 92.8 ± 5.2 | 93.0 ± 5.0 | 0.953 | 0.951 (0.91–0.97) | 0.986 (0.97–0.99) | 0.973 (0.95–0.98) | 0.72 | 2.00 | 1.63 | 0.36 |
Squat Hip Angle (°) | 78.7 ± 6.1 | 79.1 ± 6.0 | 0.952 | 0.950 (0.90–0.97) | 0.984 (0.96–0.99) | 0.971 (0.94–0.98) | 0.85 | 2.36 | 1.91 | 0.19 |
Squat Bar Velocity (m·s⁻¹) | 0.45 ± 0.07 | 0.45 ± 0.07 | 0.974 | 0.975 (0.95–0.99) | 0.994 (0.99–1.00) | 0.988 (0.98–0.99) | 0.005 | 0.014 | 0.016 | 0.57 |
Bench Elbow Angle (°) | 88.1 ± 4.8 | 88.8 ± 4.6 | 0.961 | 0.948 (0.89–0.97) | 0.985 (0.97–0.99) | 0.970 (0.94–0.98) | 0.70 | 1.94 | 1.67 | 0.003* |
Bench Bar Velocity (m·s⁻¹) | 0.30 ± 0.06 | 0.30 ± 0.06 | 0.969 | 0.968 (0.94–0.98) | 0.992 (0.98–1.00) | 0.985 (0.97–0.99) | 0.006 | 0.017 | 0.015 | 0.47 |
Deadlift Knee Angle (°) | 108.7 ± 6.3 | 108.9 ± 6.1 | 0.950 | 0.949 (0.89–0.97) | 0.984 (0.96–0.99) | 0.969 (0.94–0.98) | 0.87 | 2.41 | 1.67 | 0.48 |
Deadlift Hip Angle (°) | 73.1 ± 5.3 | 73.0 ± 5.2 | 0.946 | 0.944 (0.88–0.97) | 0.982 (0.96–0.99) | 0.967 (0.93–0.98) | 0.89 | 2.47 | 1.86 | 0.84 |
Deadlift Bar Velocity (m·s⁻¹) | 0.36 ± 0.07 | 0.36 ± 0.07 | 0.982 | 0.981 (0.96–0.99) | 0.995 (0.99–1.00) | 0.989 (0.98–1.00) | 0.005 | 0.014 | 0.015 | 0.07 |
Note: Values are mean ± SD. ICC = intraclass correlation coefficient (95% CI); SEM = standard error of measurement; MDC95 = minimal detectable change; RMSE = root mean square error. p < 0.05 indicates statistical significance.
The intra-rater and inter-rater reliability analyses demonstrated excellent repeatability of the manual video-digitization procedures across all measured variables. Intra-rater ICC values ranged from 0.982 to 0.995, while inter-rater ICC values ranged from 0.967 to 0.989. For joint-angle variables, SEM ranged from 0.70° to 0.89° and MDC95 ranged from 1.94° to 2.47°. For barbell-velocity variables, SEM ranged from 0.005 to 0.006 m·s⁻¹ and MDC95 ranged from 0.014 to 0.017 m·s⁻¹. These values indicate low measurement error and high consistency between repeated measurements.
Comparison between the proposed three-camera video-based system and the expert manual video-digitization reference method demonstrated a high level of agreement across the measured variables. Although strong Pearson correlation coefficients were observed (r = 0.946–0.982, p < 0.001), measurement agreement was primarily supported by excellent ICC values (0.944–0.981), low RMSE values, and minimal systematic bias with narrow Bland–Altman limits of agreement. The highest agreement was observed for deadlift barbell velocity (ICC = 0.981; 95% CI: 0.96–0.99), followed by squat barbell velocity (ICC = 0.975; 95% CI: 0.95–0.99). The ICC values for the joint-angle variables ranged from 0.944 to 0.951, indicating excellent consistency between the two measurement methods. Paired-samples comparisons showed no statistically significant differences between the proposed system and the expert manual reference method for seven of the eight variables (p > 0.05). The mean difference in elbow angle for bench press was statistically significant (p = 0.003), but not practically significant as it was less than 1°. The correspondence between the proposed three-camera video-based system and the expert manual reference measurements is shown in Figure 3. All plotted data points represent individual paired observations derived from the study dataset.
Measurement error remained low across all variables. RMSE values ranged from 1.63° to 1.91° for joint angles and from 0.015 to 0.016 m·s⁻¹ for barbell velocity, indicating high precision for field-based kinematic assessment.
Variable | Mean Bias | Lower LOA | Upper LOA |
|---|---|---|---|
Squat Knee Angle (°) | -0.23 | -3.42 | 2.96 |
Squat Hip Angle (°) | -0.39 | -4.13 | 3.35 |
Squat Bar Velocity (m·s⁻¹) | -0.001 | -0.032 | 0.030 |
Bench Elbow Angle (°) | -0.74 | -3.99 | 2.51 |
Bench Bar Velocity (m·s⁻¹) | 0.002 | -0.028 | 0.032 |
Deadlift Knee Angle (°) | -0.18 | -3.45 | 3.09 |
Deadlift Hip Angle (°) | 0.06 | -3.58 | 3.70 |
Deadlift Bar Velocity (m·s⁻¹) | -0.004 | -0.033 | 0.025 |
Note. Mean bias = average difference between methods; LOA = limits of agreement (mean bias ± 1.96 SD).
Small mean differences were observed between the methods, with biases generally near zero for all variables (Table 3), as determined by Bland–Altman analysis. The 95% limits of agreement were relatively narrow and generally symmetric across the measurement range, suggesting consistent agreement across the observed values. For example, the mean bias for deadlift barbell velocity was −0.004 m·s⁻¹, with limits of agreement from −0.033 to 0.025 m·s⁻¹, indicating limited variability between methods. Some wider limits of agreement were observed for hip-angle variables, especially in the squat, but these remained within a reasonable range for applied biomechanical analysis. Visual inspection of the Bland–Altman plots did not suggest clear proportional bias. Bland–Altman plots are used to illustrate agreement between the two measurement methods (Figure 4).
Figure 4. Bland–Altman plot showing agreement between the proposed three-camera video-based system and the expert manual reference method for a representative variable. Mean bias and 95% limits of agreement are indicated. The plot demonstrates minimal systematic bias and narrow 95% limits of agreement between the two measurement methods.
Overall, the proposed three-camera video-based system demonstrated a high level of agreement with the expert manual video-digitization reference method across all measured variables, together with low measurement error and minimal systematic bias. These findings indicate that the proposed system provides measurements that are consistent with expert manual analysis for the selected three-dimensional kinematic variables under the standardized field-based conditions investigated.
Discussion
The present study evaluated the agreement between a three-camera video-based system and an expert manual video-digitization reference method for three-dimensional (3D) kinematic analysis of powerlifting movements performed under standardized field-based conditions. The principal finding was that the proposed system demonstrated a high level of agreement with the expert manual reference method across all measured variables. This was supported primarily by excellent intraclass correlation coefficients, low root means square error, minimal systematic bias, and narrow Bland–Altman limits of agreement, indicating consistent measurement performance under the standardized field-based conditions investigated. These findings suggest that the proposed system produced measurements that were consistent with expert manual analysis for the selected kinematic variables under the experimental conditions investigated.
Accurate quantification of joint kinematics is central to biomechanical analysis, especially in high-load movements involving multiple joints, such as the squat, bench press, and deadlift. In the present study, joint-angle variables showed low RMSE values (less than 2°) and high ICC values (all greater than 0.94). The observed agreement is consistent with commonly used thresholds in applied motion analysis, where ICC values above 0.90 are typically considered excellent [19,20].
The observed agreement is similar to that reported in recent video-based and markerless motion-analysis studies, in which ICC values have ranged from 0.85 to 0.98 depending on the movement task and tracking method [21]. Likewise, errors of approximately 2–3° are generally considered small enough for applied joint-angle assessment in many biomechanical contexts. These findings support the use of multi-camera video-based approaches for acquiring kinematic data with adequate accuracy for applied performance-analysis settings outside the laboratory [22,23].
Barbell-velocity measurements showed the strongest agreement between methods, with ICC values close to 1.00 and RMSE values below 0.02 m·s⁻¹. This level of agreement is comparable with previously reported velocity-based measurement systems used in resistance training, including linear position transducers and video-based systems [24]. The high agreement for velocity measures may be partly explained by the relative ease of tracking the barbell compared with anatomical landmarks, which are more affected by soft-tissue artefact and partial occlusion. Previous work has also shown that external object tracking may produce smaller errors than joint-based kinematic estimation in video-analysis systems [25,26]. These results support the potential use of the proposed system in velocity-based training contexts where small changes in barbell velocity may inform performance and fatigue monitoring.
Bland–Altman analysis showed small mean differences and relatively narrow limits of agreement between the two measurement methods across the analyzed variables. Visual inspection did not suggest clear proportional bias, supporting consistent agreement under the tested conditions. The magnitude of measurement error observed in this study is comparable with previous comparisons between video-based and motion-capture methods, where errors are commonly influenced by landmark visibility, camera resolution, calibration procedures, and occlusion [27,28]. The slightly wider limits of agreement for hip-angle variables may be related to anatomical occlusion or trunk position during loaded movements, as previously noted in resistance-training biomechanics studies [29].
The only statistically significant difference was observed for the bench press elbow angle. Although this difference reached statistical significance, the corresponding RMSE remained low and the ICC indicated excellent agreement, suggesting that the practical magnitude of the difference was small. This difference may be attributed to the greater complexity of upper-limb kinematics during the bench press, where partial occlusion of the elbow by the barbell or upper limb, together with subtle differences in manual landmark identification, may influence angle estimation. These factors are recognised challenges in video-based biomechanical analysis and may explain the small systematic difference observed for this variable.
The present findings should not be interpreted as demonstrating equivalence with laboratory-grade optical motion-capture systems because no such criterion system was available for comparison in the present investigation. Instead, the results indicate that the proposed three-camera video-based system produced measurements that closely agreed with expert manual video digitization under standardized field-based conditions. Therefore, the findings support the potential application of the proposed system as a practical field-based biomechanical assessment tool when laboratory motion-capture facilities are unavailable, while further validation against laboratory-grade optical motion-capture systems remains warranted. It should also be recognized that expert manual video digitization is not free from measurement error and may be influenced by factors such as landmark identification, image quality, and observer judgment. Although the use of two experienced analysts, standardized digitization procedures, and excellent intra- and inter-rater reliability minimized these sources of error, the reference method should be regarded as a practical field-based comparator rather than a perfect criterion standard.
The present study extends the existing literature by evaluating multi-camera video-based motion analysis during high-load, multi-plane resistance exercises under ecologically valid conditions. Previous agreement and validation studies have reported acceptable measurement performance for gait, functional movements, and low-load resistance tasks [30–32]. The present findings suggest that comparable measurement agreement may be achievable in more complex resistance exercises when camera views are synchronized and the measurement volume is calibrated. Multi-camera approaches may reduce perspective-related error and provide more spatial information than single-camera approaches [33]. These findings are consistent with recent developments in applied biomechanics, where multi-view reconstruction techniques are increasingly used to bridge laboratory-based and field-based motion analysis [34].
Practical Implications
The findings support the practical use of low-cost multi-camera video-based systems for applied biomechanical assessment in strength-training contexts. Quantifying joint kinematics and barbell velocity under realistic training conditions may assist technique evaluation, athlete monitoring, and data-informed coaching. The portability and accessibility of the system may also facilitate biomechanical analysis outside laboratory environments. The use of a standardized load corresponding to 80% of each participant's one-repetition maximum (1RM) minimized variability associated with loading intensity and provided consistent conditions for comparing kinematic measurements across participants.
Limitations
The present study has several limitations. The participants consisted only of young male university powerlifters from a single weight category, which may limit the generalizability of the findings. In addition, the proposed three-camera video-based system was compared with an expert manual video-digitization reference method rather than a laboratory-grade optical motion-capture system. Quantitative calibration error, reprojection error, and reconstruction accuracy were not evaluated, and only one trial was analysed for each lift, precluding assessment of trial-to-trial variability and within-session repeatability. Therefore, the findings should be interpreted as evidence of agreement with an expert manual reference method under the specific experimental conditions investigated rather than as criterion validation against laboratory motion-capture technology. Future studies should evaluate the performance of the proposed system across a wider range of loading intensities, including near-maximal and maximal lifts, to determine its applicability under different resistance-training conditions.
Conclusions
The present study demonstrated that the proposed three-camera video-based system showed high agreement with an expert manual video-digitization reference method for measuring selected three-dimensional kinematic variables during squat, bench press, and deadlift exercises under standardized field-based conditions. Although the system was not compared with a laboratory-grade optical motion-capture system, the findings suggest that it may provide a practical option for field-based biomechanical assessment when laboratory facilities are unavailable. Further validation against established laboratory motion-capture systems is recommended.
Author Contributions
Conceptualization, SD and RR; Methodology, SD and RR; Software, SD; Validation, SD and RR; Formal analysis, SD; Investigation, SD and RR; Resources, RR; Data curation, SD; Writing – original draft, SD; Writing – review and editing, RR; Visualization, SD; Supervision, RR; Project administration, RR; Funding acquisition, RR. All authors have read and approved the final version of the manuscript.
Funding
This research received no external funding.
Ethical Consideration
Ethical review and approval were obtained from the Institutional Human Research Ethics Committee of the host institution (Approval No. HITS/DRC/22701008/2026) in accordance with the Declaration of Helsinki.
Informed Consent Statement
Written informed consent was obtained from all participants before participation. Participants were informed of the study objectives, procedures, and potential risks.
Data Availability Statement
The data analyzed in the present study are available from the corresponding author upon reasonable request, subject to data-protection and ethical regulations.
Acknowledgments
The authors thank all the participants for their time and dedication in the study.
Conflicts of Interest
The authors declare no conflicts of interest.
References
- Khuyagbaatar, B.; Tumurbaatar, M.; Tsenkherjav, K.; Purevsuren, T.; Shambaljamts, T.; Kim, K.; Danjkhuu, T.; Ganbat, D.; Kim, Y.H. Kinematic Comparison of Snatch and Clean Lifts in Weightlifters Using Wearable Inertial Measurement Unit Sensors. Physical Activity and Health 2024, 8, 1–9, doi:10.5334/paah.306.
- Rodal, M.; Franco-García, J.M.; Carlos‐Vivas, J.; José, F.; Gosálvez, A.P.; Charneco-Salguero, G.; Gianikellis, K.; Pérez‐Gómez, J. Spatio-Temporal and Mechanical Analysis of Bench Press Phases: Barbell Kinematics and Dynamics Across Different Load Intensities. Applied Sciences 2024, 14, 11783–11783, doi:10.3390/app142411783.
- Sandau, I.; Langen, G.; Nitzsche, N. Variability of Time Series Barbell Kinematics in Elite Male Weightlifters. Frontiers in Sports and Active Living 2023,5, doi:10.3389/fspor.2023.1264280.
- Nakano, N.; Sakura, T.; Ueda, K.; Omura, L.; Kimura, A.; Iino, Y.; Fukashiro, S.; Yoshioka, S. Evaluation of 3D Markerless Motion Capture Accuracy Using OpenPose with Multiple Video Cameras. Frontiers in Sports and Active Living. 2020, 2, 50. doi 10.3389/fspor.2020.00050
- Turner, J.A.; Chaaban, C.R.; Padua, D.A. A low-cost markerless motion capture system for lower-extremity kinematics during return-to-sport tasks. Journal of Biomechanics. 2024; 171:112200. https://doi.org/10.1016/j.jbiomech.2024.112200
- Hamilton, R.I.; Williams, J.; Holt, C.A. Biomechanics beyond the Lab: Remote Technology for Osteoarthritis Patient Data—A Scoping Review. Frontiers in Rehabilitation Sciences 2022,3, doi:10.3389/fresc.2022.1005000.
- Kakavand, R.; Ahmadi, R.; Parsaei, A.; Edwards, W.B.; Komeili, A. Comparison of kinematics and kinetics between OpenCap and a marker-based motion capture system in cycling. Computers in Biology and Medicine. 2025;192(Pt A):110295. doi 10.1016/j.compbiomed.2025.110295.
- Stamm, O.; Steinert, A. Accuracy of Monocular Two-Dimensional Pose Estimation Compared with a Reference Standard for Kinematic Multiview Analysis: Validation Study. JMIR mHealth and uHealth. 2020, 8, e19608. doi 10.2196/19608
- Cronin, N.J.; Walker, J.; Tucker, C.B.; Nicholson, G.; Cooke, M.P.; Merlino, S.; Bissas, A. Feasibility of OpenPose Markerless Motion Analysis in a Real Athletics Competition. Frontiers in Sports and Active Living 2024, 5, doi:10.3389/fspor.2023.1298003.
- Uhlrich, S.D.; Falisse, A.; Kidziński, Ł.; Muccini, J.; Ko, M.; Chaudhari, A.S.; Hicks, J.L.; Delp, S.L. OpenCap: Human Movement Dynamics from Smartphone Videos. PLoS Computational Biology. 2023, 19(10), e1011462. doi 10.1371/journal.pcbi.1011462
- Thiele, F.; Paternoster, F.K.; Hummel, C.; Stöcker, F.; Holzer, D. Assessment of the Accuracy of a Deep Learning Algorithm- and Video-Based Motion Capture System in Estimating Snatch Kinematics. International Journal of Exercise Science. 2024, 17(1), 1629–1647. https://doi.org/10.70252/PRVV4165.
- Brice, S.M.; Doma, K.; Spratford, W. Effect of Footwear on the Biomechanics of Loaded Back Squats to Volitional Exhaustion in Skilled Lifters. The Journal of Strength and Conditioning Research 2021, 36, 2676–2684, doi:10.1519/jsc.0000000000003986.
- Shishov, N.; Elabd, K.; Komisar, V.; Chong, H.; Robinovitch, S.N. Accuracy of Kinovea Software in Estimating Body Segment Movements during Falls Captured on Standard Video: Effects of Fall Direction, Camera Perspective and Video Calibration Technique. PLoS ONE 2021, 16, doi: 10.1371/journal.pone.0258923.
- Bolstad, P.V.; Gessler, P.E.; Lillesand, T.M. Positional Uncertainty in Manually Digitized Map Data. International Journal of Geographical Information Systems 1990, 4, 399–412, doi:10.1080/02693799008941555.
- Puig-Diví, A.; Escalona‐Marfil, C.; Riu, J.M.; Busquets, A.; Padullés-Chando, X.; Marcos-Ruiz, D. Validity and Reliability of the Kinovea Program in Obtaining Angles and Distances Using Coordinates in 4 Perspectives. PLoS ONE 2019,14, doi: 10.1371/journal.pone.0216448.
- Alonso-Aubín, D.A.; Chulvi‐Medrano, I.; Cortell-Tormo, J.M.; Picón-Martínez, M.; Rebullido, T.R.; Faigenbaum, A.D. Squat and Bench Press Force-Velocity Profiling in Male and Female Adolescent Rugby Players. The Journal of Strength and Conditioning Research 2019, 35, doi:10.1519/jsc.0000000000003336.
- Hani, H.; Souchereau, R.; Kachlan, A.; Harris, H.; Dufour, J.; Aurand, A.; Mageswaran, P.; Hyer, M.; Marras, W.S. Reliability of a Wearable Motion System for Clinical Evaluation of Dynamic Lumbar Spine Function. Advances in Complementary & Alternative Medicine. 2022, 7(2), 672–683. https://doi.org/10.31031/ACAM.2022.07.000660.
- Koo, T.K.; Li, M.Y. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. Journal of Chiropractic Medicine. 2016;15(2):155–163. doi 10.1016/j.jcm.2016.02.012
- Wang, X.M.; Smith, D.T.; Zhu, Q. A Webcam-Based Machine Learning Approach for Three-Dimensional Range of Motion Evaluation. PLoS ONE 2023, 18, doi: 10.1371/journal.pone.0293178.
- Philipp, N.M.; Cabarkapa, D.; Cabarkapa, D.V.; Eserhaut, D.A.; Fry, A.C. Inter-Device Reliability of a Three-Dimensional Markerless Motion Capture System Quantifying Elementary Movement Patterns in Humans. Journal of Functional Morphology and Kinesiology. 2023, 8(2), 69. doi 10.3390/jfmk8020069
- Lima, Y.L.; Donnelly, C.J.; Uhlrich, S.D.; Falisse, A.; Delp, S.L.; Pappas, E.; Mentiplay, B.F. Validity and reliability of trunk and lower-limb kinematics measured using OpenCap during functional movements. Journal of Sports Sciences. 2024. doi 10.1080/02640414.2024.2415233.
- Needham, L.; Evans, M.; Wade, L.; Cosker, D.P.; McGuigan, P.M.; Bilzon, J.L.; Colyer, S.L. The Development and Evaluation of a Fully Automated Markerless Motion Capture Workflow. Journal of Biomechanics. 2022, 144, 111338. doi 10.1016/j.jbiomech.2022.111338.
- Wade, L.; Needham, L.; Evans, M.; McGuigan, P.; Colyer, S.; Cosker, D.; Bilzon, J. Examination of 2D Frontal and Sagittal Markerless Motion Capture: Implications for Markerless Applications. PLoS ONE 2023, 18, doi: 10.1371/journal.pone.0293917.
- Courel‐Ibáñez, J.; Martínez‐Cava, A.; Morán‐Navarro, R.; Escribano‐Peñas, P.; Chavarren-Cabrero, J.; González‐Badillo, J.J.; Pallarés, J.G. Reproducibility and Repeatability of Five Different Technologies for Bar Velocity Measurement in Resistance Training. Annals of Biomedical Engineering 2019, 47, 1523–1538, doi:10.1007/s10439-019-02265-6.
- Koul, A.; Novembre, G. How Accurately Can We Estimate Spontaneous Body Kinematics from Video Recordings? Effect of Movement Amplitude on OpenPose Accuracy. Behavior Research Methods 2025, 57, 38–38, doi:10.3758/s13428-024-02546-6.
- Cunanan, A.J.; Hornsby, W.G.; South, M.A.; Ushakova, K.P.; Mizuguchi, S.; Sato, K.; Pierce, K.C.; Stone, M.H. Survey of Barbell Trajectory and Kinematics of the Snatch Lift from the 2015 World and 2017 Pan-American Weightlifting Championships. Sports 2020, 8, 118–118, doi:10.3390/sports8090118.
- Conconi, M.; Pompili, A.; Sancisi, N.; Parenti‐Castelli, V. Quantification of the Errors Associated with Marker Occlusion in Stereophotogrammetric Systems and Implications on Gait Analysis. Journal of Biomechanics 2020, 114, 110162–110162, doi: 10.1016/j.jbiomech.2020.110162.
- Mata, A.J.; Hayashi, H.; Moreno, P.A.; Dudley, R.I.; Sorenson, E.A. Hip Flexion Angles During Supine Range of Motion and Bodyweight Squats. International Journal of Exercise Science. 2021, 14(1), 912–918. https://doi.org/10.70252/LMEG8374.
- Straub, R.K.; Powers, C.M. A Biomechanical Review of the Squat Exercise: Implications for Clinical Practice. International Journal of Sports Physical Therapy. 2024, 19(4), 490–501. doi 10.26603/001c.94600
- Bissas, A.; Cronin, N.J. Evaluating Markerless Biomechanical Analysis in a Real-World Pole Vault Competition Setting. PLOS ONE. 2026, 21(3), e0329987. doi 10.1371/journal.pone.0329987
- Colyer, S.L.; Evans, M.; Cosker, D.P.; Salo, A.I.T. A Review of the Evolution of Vision-Based Motion Analysis and the Integration of Advanced Computer Vision Methods Towards Developing a Markerless System. Sports Medicine - Open. 2018, 4, 24. doi 10.1186/s40798-018-0139-y
- Horsak, B.; Prock, K.; Krondorfer, P.; Siragy, T.; Simonlehner, M.; Dumphart, B. Inter-trial variability is higher in 3D markerless compared to marker-based motion capture: Implications for data post-processing and analysis. Journal of Biomechanics. 2024;166:112049. https://doi.org/10.1016/j.jbiomech.2024.112049
- Klimaszewski, K.; Grajek, T.; Wegner, K. Factors Influencing Accuracy of Estimating Position of Objects in a Multi-Camera System. Journal of Telecommunications and Information Technology 2024, 1–16, doi:10.26636/jtit.2024.2.1457.
- Das, K.; de Paula Oliveira, T.; Newell, J. Comparison of markerless and marker-based motion capture systems using 95% functional limits of agreement in a linear mixed-effects modelling framework. Scientific Reports. 2023, 13, 22880. https://doi.org/10.1038/s41598-023-49360-2.
