Elsevier

Journal of Biomechanics

Volume 41, Issue 12, 28 August 2008, Pages 2776-2780
Journal of Biomechanics

Short communication
Systematic accuracy and precision analysis of video motion capturing systems—exemplified on the Vicon-460 system

https://doi.org/10.1016/j.jbiomech.2008.06.024Get rights and content

Abstract

With rising demand on highly accurate acquisition of small motion the use of video-based motion capturing becomes more and more popular. However, the performance of these systems strongly depends on a variety of influencing factors. A method was developed in order to systematically assess accuracy and precision of motion capturing systems with regard to influential system parameters.

A calibration and measurement robot was designed to perform a repeatable dynamic calibration and to determine the resultant system accuracy and precision in a control volume investigating small motion magnitudes (180×180×150 mm3). The procedure was exemplified on the Vicon-460 system. Following parameters were analyzed: Camera setup, calibration volume, marker size and lens filter application.

Equipped with four cameras the Vicon-460 system provided an overall accuracy of 63±5 μm and overall precision (noise level) of 15 μm for the most favorable parameter setting. Arbitrary changes in camera arrangement revealed variations in mean accuracy between 76 and 129 μm. The noise level normal to the cameras’ projection plane was found higher compared to the other coordinate directions. Measurements including regions unaffected by the dynamic calibration reflected considerably lower accuracy (221±79 μm). Lager marker diameters led to higher accuracy and precision. Accuracy dropped significantly when using an optical lens filter.

This study revealed significant influence of the system environment on the performance of video-based motion capturing systems. With careful configuration, optical motion capturing provides a powerful measuring opportunity for the majority of biomechanical applications.

Introduction

Optoelectronic motion capturing is widely used in biomechanics. Applications range from macroscopic gait analysis (Blake and Ferguson, 1993) to measurement of implant micromotion (Lechert et al., 2003). The theoretical background, deriving 3D coordinates of retroreflective markers from several 2D camera projections, has been sufficiently established and extensively applied (Abdel-Aziz and Karara, 1971; Chen et al., 1994). Performance of motion capturing systems strongly depends on their setup and is highly sensitive against alterations. Marker properties, optical projections, video-digital conversion, camera configuration, lens distortion, calibration procedure, etc. influence the performance to various extents (Furnée, 1991). Producers state only rough estimates regarding accuracy and precision of their products. The literature reveals investigations only related to specific experimental setups (Ehara, 2002; Kwon and Casebolt, 2006; Liu et al., 2007). The purpose of the study was to develop and apply a method to systematically assess the accuracy and precision of motion capturing systems with respect to the setting of several system parameters in a clearly defined working space.

Section snippets

Materials and methods

A calibration and measurement robot was developed to achieve a repeatable dynamic calibration simultaneously with a semi-automatic accuracy and precision analysis. Besides carrying out arbitrarily defined wand motion for calibration, the robot was capable of moving a marker successively to predefined gridpoints to analyze the resulting system performance (grid measurement). The procedure was exemplified on the Vicon-460 system (Vicon Motion Systems, LA, USA) using MCam-60 cameras and the

Results

An overall accuracy of 63±5 μm (precision 15 μm) was found for the most favorable parameter combination (Camera setup: 2, Calibration: “full”, Marker size: 25 mm, Lens filter: without).

Camera setup significantly influenced the overall accuracy (ANOVA 1, p<0.001; Table 2). Highest accuracy was observed for setup 2 (76±3 μm, mean±SD). Lowest accuracy (129±35 μm) was found for setup 4 (Fig. 4). Maximum observed gridpoint error was 416±129 μm (setup 4). Statistical subsets were identified for setups 1–3

Discussion

This study revealed significant impact of several parameters on the performance of the Vicon-460 system. The analysis was carried out within an observation volume as typically used for biomechanical applications investigating small motion magnitudes. Normalizing the results to the volume size might provide an estimate for other volumes. It is assumed that at least tendencies of the outcomes are valid for the majority of motion capturing systems with equal underlying theoretical principles. In

Conflict of interest statement

We do not have any proprietary, financial, professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, this manuscript.

Acknowledgments

We would like to thank Dr. Mathias Bankay, Ms. Irène Vollmer, and Mr. Martin-Scott Löhrer from prophysics AG—German/Swiss Vicon Distributor, for their assistance and support.

References (9)

There are more references available in the full text version of this article.

Cited by (299)

  • A Data-Driven Model for Range Sensors

    2024, International Journal of Semantic Computing
View all citing articles on Scopus
View full text