3-DIMENSIONAL VIDEOGRAPHY OF SWIMMING WITH FANNING PERISCOPES

Citation
T. Yanai et al., 3-DIMENSIONAL VIDEOGRAPHY OF SWIMMING WITH FANNING PERISCOPES, Journal of biomechanics, 29(5), 1996, pp. 673-678
Citations number
8
Categorie Soggetti
Engineering, Biomedical",Biophysics
Journal title
ISSN journal
00219290
Volume
29
Issue
5
Year of publication
1996
Pages
673 - 678
Database
ISI
SICI code
0021-9290(1996)29:5<673:3VOSWF>2.0.ZU;2-P
Abstract
The purpose of this paper is to describe a three-dimensional videograp hy method with panning periscopes for reconstructing swimming techniqu es. Two panning periscope systems, developed in our laboratory, were u sed for the data collection. A control object (1 x 1.2 x 2) m(3) with 72 markers was placed in seven consecutive locations along the (1.5 x 8.4 x 2) m(3) calibration space. The position of the control object wa s recorded while panning the periscope systems, and for each of the se ven locations the DLT camera parameters were determined for eight fiel ds equally spaced through the panning motion. Each DLT camera paramete r was then expressed as a function of the camera orientation by using a least-squares, second-order, polynomial regression equation, so that the DLT camera parameters for any camera orientation could be predict ed (Yu et al., J. Biomechanics 26, 741-751, 1993). The three-dimension al coordinates of desired body landmarks at an instant were determined from the digitized two-dimensional coordinates for two cameras and th e predicted DLT camera parameters at that instant. The accuracy of the method was evaluated by (a) the resultant errors in computing three-d imensional coordinates of precalibrated static points, and (b) the err ors in computing the length of a scale rod pulled through the calibrat ion space. The mean resultant errors ranged from 8.34 to 16.44 mm for the above- and from 9.93 to 16.22 mm for the below-water control volum es. The mean error in computing lengths on the scale rod ranged from 3 .32 to 5.83 mm for three 0.5 m lengths, and 9.97 mm for a 1.5 m length . The method produced acceptable results in the reconstruction of thre e-dimensional motions recorded from a large space above and below the water surface.