id thus 
time. 
art 
lage 
ata 
  
  
itor 
ata 
igle 
Image 
itor 
jan gait 
le video 
  
  
Figure 2 shows the HVT system and test field of 
gait experiment, and experiments procedure are 
as follows: 
+ The subject walked along the 10m course 
under the condition that the right knee and 
right ankle was fixed by brace on the 
assumption of suffer. 
+ In order to acquire the 3D coordinate of 
subject's feature points such as the head, 
shoulder, knee or so on, markers were fitted on 
the subject's body. 
* The subject was illuminated by the two video 
lights mounted on the pan head, and aperture 
of the cameras were turned down for effective 
marker extraction. 
CCD 3 camera mounted on center of the pan head 
is connected to PC directly, then image processing 
for tracking are performed in real-time, and the 
rotation of the HVT system can be controlled using 
the results of image processing. The authors 
developed an auto tracking function for moving 
object using template matching based on SSDA 
method in this paper. Figure 3 shows the basic 
flow of the auto tracking procedure for the HVT 
system, and detail procedures are as follows: 
1. Camera parameters (focal length, scale factor) 
of the CCD 3 are inputted 
2. The HVT is rotated to the start position and 
image for the subject on the start position is 
taken. Also, template area and search area are 
determined. 
3. Subject start gait motion. 
4. Tracking image is taken. 
5.Template matching is performed, and image 
coordinate forthe center of template is obtained. 
6. Changing values for horizontal and vertical 
angle are calculated from image coordinate and 
camera parameters. 
7. Repeating the tracking processing. 
In order to increasing the tracking speed, the 
resolution of tracking image was compressed 
asl50 Pixel (H) - 120 Pixel (VJ) (8 bit 
monochrome) in this paper. Figure 4 shows the 
tracking image of this experiment, and the 
maximum speed of tracking was 12 frames per 
second. 
4 GAIT ANALYSIS 
The analysis of walking cycle is effective method 
walking cycle is divided into two phases, i.e., stance 
walking elements. 
Stance phase Swing phase 
+ Heel strike +Acceleration 
+ Foot flat +Mid swing 
+ Mid stance +Deceleration 
+ Heel off 
+ Toe off 
  
  
   
  
Anai, Tetsuji 
  
  
g image 
  
  
Image Processing 
(SSDA Jodi 
WE Image coordinates B 
H.V. Angle 
Lan MS 
Yes 
End 
  
  
  
  
  
Figure 3. Basic Flow of Tracking 
3 
Template Area 
Search Area 
i 
Figure 4. Tracking Image 
for estimation of human gait motion. Generally, Human 
phase and swing phase, and each phase consist of following 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 27