are 
t by 
The 
(2) 
tion 
phi- 
and 
and 
ı the 
and 
-hand 
rior 
mera 
1ined 
here 
phi 
each 
(cept 
   
the goal and start images, it’s 
impossible to estimate X-coordinate(along 
track). However, Y-coordinate(cross 
track) can be estimated from the width of 
track line. 
Although initial kappa can be regarded as 
a line inclination angle, omega, phi and 
kappa can be calibrated from the second 
equation of Equ. (1) and (23. 
Finally, exchanging Y and Z axis under 
the condition that Carl Lowis is running 
on the same parallel plane(2) 
perpendicular to the camera (Fig.10), 
feature points coordinates (X'.Y')" are 
calculated from Equ. (3). 
a11X+a91Y-a31f 
X' »(Z'-Z29') ————Fm—— + Xo’ 
813X*853Y-833f 
(3) 
a19X+aggy-agof 
Y'-(Z og!) — —— ——— « Yg! 
813X*853Y-a33f 
in Fqu.(3), (257291) is constant as a 
distance from the camera to the running 
track of Carl Lowis is constant. 
Z(Y) 
Y(Z) 
  
  
  
i Constant 
É 
Camera Position 
Fig.10 Geometric relationship between 
XYZ and X'Y'Z*' planes 
5. Conclusions 
An image processing procedure for 
feature extraction of human body has 
been developed in this study. 13 feature 
points are automatically extracted by 
this procedure with 32 bit PC. Although 
binary image which is taken through the 
image enhancement process is an image 
like silhouette, it's impossible to 
extract all feature points automatically. 
However, dynamic analysis for some 
feature points for example knee can be 
performed by this procedure. The 
horizontal displacement of Carl Lowis's 
right knee was analyzed in this study. 
Orientation for each image should be 
performed by orientation method which is 
discussed in this paper. 
Consequently, this method should be 
improved further. For example, on the way 
of image enhancement process, binary 
image for only white track line should be 
created. This image will give useful 
information for orientation such as a 
line inclination(kappa), width between 
the track lines and pixel-line 
coordinates for Cross points with 
horizontal and vertical lines. These 
automated process should be developed for 
dynamic analyses with sequential multi- 
images of TV video camera. 
Reference 
1)H.Chikatsu; Simultaneous Adjustment for 
a self-Calibration of a CCD Camera in a 
2D Field. 1st International Symposium on 
Application of Geodesy to Engineering, 
Stuttgart, May 1991. 
2)Y.D.Huang and I.Harley; Camera 
Calibration without a Control Field - a 
New Method, Optical 3-D Measurement 
Techniques. pp.49-56, Wien, Sept,1989. 
3)G.Bayer,P.Krzystekand,W.Móhhlenbrink; 
Real Time Positioning of Moving Objects 
by Dynamic Target Tracking, 16th  ISPRS 
KYOTO CONGRESS, COMMISSION 5, pp.32-43, 
Kyoto, Aug. 1988.