1eck points
SES
sure station
in by rotating the video
nd 27 ° 46'02" above the
nd AH,=0°0"17".
1 by rotating the video
and 27^ 43'52" above the
nd AH,z27'43'52".
yr each sequential image
of changing vertical and
,and ¢ , respectively.
libration results of the
, wand ¢ are calculated
dues in vertical (A V) ,
Do.
"y"and AH >0 in the
tion between telescope
dolite coordinate system
was taken at a different
in figure 8, due to the
e theodolite and the lens
era position has to be
rotation of the video
on,
/ienna 1996
x - D'cosV sin (q - Ho)
Yo = of cos Sin V. - sino cosV cos( - D (3)
of
Z - DÀ sino sin V + coso cosV cos(,- H,
X pg
Z = (p-z) cos Vo
where,X,*, Y ,*,Z,*; corrected camera position,
D'-A/ x8 « YB «(p —zoJ- , v" z tan (Yo / (D - z)))
Then, the X and Y coordinates for checking points in
each sequential image are computed from the following
equation,
. ax+a,y-a.,f *
X2Xpg A TT m4 31 Z—2)
aX t Ay — af (4)
. àax+a.y-at 2
Y.= Yo + EX Ta Al 2 % Z-2p)
aX + a,y —a,f
where, X, Y,Z ; object coordinate, x,y ; image coordinate, f
; focal length, a ;;rotation matrix with three parameters,
0,0, kK.
Table 3 shows the RMSE for check points on the target
field B and C by using calibration parameters which were
acquired as noted above.
Consequently, utilization of the video theodolite system
is expected to become a useful tool in the field of sports
training and rehabilitation since the camera parameters
can be acquired in real time and the position for feature
points of a human can be calculated.
4. APPLICATION OF VIDEO THEODOLITE
The relatively slow 4 images acquired per second is
perhaps due to the ability of image processing board or
MET2NV. Test for dynamic analysis of human motion were
performed in a gymnasium by using MET2NV( Figure 9).
Figure 9. Test field for dynamic analysis
Sequential images were taken from 4.5 m at the above
noted intervals. Calibration was performed by using one
sheet reflector fitted on the wall. Table 4 shows the
calibration results with 9 control points.
The authors have previously analyzed the dynamics of
the sprinter Carl Lewis and that of boat rowing by using an
image procedure and animation technique(Chikatsu and
Murai,1992,1994). For the automated recognition of some
feature points of a human, the image processing
procedure and animation technique were combined in this
paper from the above experience. The most remarkable
point of this automated recognition is template matching.
Templates have coordinates for some feature points of a
human in advance. Then, only template matching between
original binary image and template binary image are
needed. Feature points of a human are shown in figure
10.
Figure 10. Feature points
Figure 12. Animated cartoons
Table 4. Calibration results
f^ exo v Zi o ó K
mm mm
mm
-0.358 | -55.991 4496.293 —0°5'50”" —-0°22'43"- 0°9'18" 277.998 257.249 5.938 131.792 -0.070 0.412
X, Yo f a, &
pixel pixel mm
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
