Sarfaty O, Ladin Z
Biomedical Engineering Department, Boston University, MA 02215.
J Biomech. 1993 Aug;26(8):1011-6. doi: 10.1016/0021-9290(93)90061-i.
A system for the estimation of the inertial properties of human body segments using advanced video technology and computer image processing was developed. The system is based on the photogrammetric technique, where three-dimensional information is determined from two separate two-dimensional video images. The inertial properties are calculated using an image-processing algorithm which provides volumetric information, coupled with a database of anatomical densities provided in the literature. In order to determine the accuracy of the system and its limitations, the system estimates of the inertial properties of solid bodies were compared to theoretically calculated values. The application of the system to kinesiological studies is illustrated by measuring the inertial properties of the shank of three subjects, and comparing the results to data generated using regression equations provided in the literature. Human factors, such as segment boundaries identification and color thresholds selection, were found to introduce the largest errors. A proper selection of the optical setting can reduce the errors to levels of 5% or better. On the average, the system overestimated the inertial properties of solid objects by 2.51% for mass, 1.21% for center of mass, 4.53% for transverse moments of inertia and 3.65% for longitudinal moment of inertia. The video-based estimates of the mass and center of mass of the shank were comparable to values obtained from anthropometric-based regression equations. The predictions of the transverse moment of inertia of the shank varied considerably among the methods. The findings suggest that a video-based system represents a promising technique for estimating inertial properties of human body segments for individual subjects.(ABSTRACT TRUNCATED AT 250 WORDS)
开发了一种使用先进视频技术和计算机图像处理来估计人体各部分惯性特性的系统。该系统基于摄影测量技术，从两个单独的二维视频图像中确定三维信息。惯性特性是使用一种提供体积信息的图像处理算法，并结合文献中提供的解剖密度数据库来计算的。为了确定该系统的准确性及其局限性，将该系统对固体物体惯性特性的估计与理论计算值进行了比较。通过测量三名受试者小腿的惯性特性，并将结果与使用文献中提供的回归方程生成的数据进行比较，说明了该系统在运动学研究中的应用。发现诸如节段边界识别和颜色阈值选择等人为因素会引入最大的误差。适当选择光学设置可以将误差降低到5%或更低的水平。平均而言，该系统对固体物体惯性特性的高估情况为：质量高估2.51%，质心高估1.21%，横向转动惯量高估4.53%，纵向转动惯量高估3.65%。基于视频对小腿质量和质心的估计与从基于人体测量学的回归方程获得的值相当。不同方法对小腿横向转动惯量的预测差异很大。研究结果表明，基于视频的系统是一种很有前景的技术，可用于估计个体受试者人体各部分的惯性特性。（摘要截短为250字）
