z[m] Person # 
—17 
— 19 
  
  
  
  
  
  
  
  
  
  
  
  
x[m] 
T 
ae 
-8.0 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 
Figure 8. Passenger flow information 
Table 9. Ticket gate choice with origin information 
  
A B C D E Total 
  
North exit | 12/12 | 1/3 0/0 0/0 N/A 13/15 
  
South exit | 0/0 5/5 3/4 4/4 N/A 12/13 
  
Platform 1 | N/A 0/0 0/0 0/0 0/0 0/0 
  
Platform 2 | N/A 0/0 0/1 7/7 5/6 12/14 
  
Total 12/2 | 6/8 3/5 13/11 | 3/6 37/42 
  
  
  
  
  
  
  
  
  
Result of automatic / manual acquisition. N/A means one-way 
gate. Row is origin, column is chosen gate. Gates are named 
from A to E, respectively from left to right on figure 5. 
Table 10. OD matrix 
  
  
  
  
  
  
North South #1 #2 Total 
North exit E 0/0 1/2 3/11 4/13 
South exit 1/1 m 1/1 11/11 13/13 
Platform 1 0/0 0/0 ges 0/0 0/0 
Platform 2 3/3 6/10 0/0 zm 9/13 
Total 4/4 6/10 2/3 14/22 26/39 
  
  
  
  
  
  
  
  
Result of automatic / manual acquisition. Row is origin, column 
is destination. 
by observation model p(z,;|x,;), the same as tracking. This 
simple integration brings good result in plain situation like 
without occlusion and proximity. We need additional step to 
deal with complicated situation. 
6. CONCLUSION 
We propose a new method to track multiple human in complex 
situations. We assume human tracking as data assimilation and 
combine observed information of color and range with 
pedestrian behavior model in general state space model. From 
some applications, we show the high performance of proposed 
method. We also show the acquisition of passenger flow 
   
information using tracking result. It is expected that enormous 
human choices in the real situation will be offered, for 
automatic tracking can deal with much amount of data. 
Proposed method can be easily applied to other Situations. 
According to observation sites and human behavior there, we 
can use different pedestrian behavior model by replacing system 
model. In the same way, we can introduce different sensors such 
as range scanner and infrared sensor by replacing observation 
vector and observation model. 
Further works are as follows. Firstly, we need to make better the 
components of general state space model defined in section 3 
for more accurate tracking. Secondly, automatic human 
detection is necessary to achieve the long time tracking. For this 
problem, simple framework is already completed as explained 
in 5.5, so their expansion is the next work. Furthermore, we aim 
to develop a method to analyze pedestrian behavior using 
tracking results. 
References: 
Ali, I. and Dailey, M. 2009. Multiple human tracking in high- 
density crowds, Advanced Concepts for Intelligent Vision 
Systems, Vol. LNCS 5807, pp.540-549. 
Bierlaire, M. and Robin, T. 2009. Pedestrian choices, In H. 
Timmermans (Ed.) Pedestrian Behavior: Models, Data 
Collection and Applications, pp.1-26, Emerald Group, 2009. 
Daley, R. 1991. Atmospheric Data Analysis, Cambridge 
University Press, Cambridge. 
Gordon, N. J., Salmond, D. J. and Smith, A. F. M. 1993. Novel 
approach to nonlinear / non-Gaussian Bayesian state estimation, 
Radar and Signal Processing, IEE Proceedings F, Vol. 140, 
No.2, pp.107-113. 
Higuchi, T. 2003. Data assimilation with Monte Carlo mixture 
Kalman filter toward space weather forecasting, Proceedings of 
International Symposium on Information Science and Electrical 
Engineering 2003, pp.122-125. 
Munoz-Salinas, R. 2008. A Bayesian plan-view map based 
approach for multiple-person detection and tracking, Pattern 
Recognition, Vol.41, No.12, pp.3665-3676. 
Robin, T., Antonini, G., Bierlaire, M. and Cruz, J., 2009. 
Specification, estimation and validation of a pedestrian walking 
behavior model, Transportation Research Part B: 
Methodological, Vol.43, No.1, pp.36-56. 
Wu, B. and Nevatia, R. 2007. Detection and tracking of 
multiple, partially occluded humans by bayesian combination of 
edgelet based part detectors, International Journal of Computer 
Vision, Vol.75, No.2, pp.247-266. 
Wunsch, C. 1996. The Ocean Circulation Inverse Problem, 
Cambridge University Press, Cambridge. 
ACKNOWLEDGEMENTS: 
This research was supported by Japan Society for the Promotion 
of Science, Grant-in-Aid for Young Scientists B (22760401). 
   
  
     
     
   
  
    
   
  
   
   
     
   
      
        
      
       
          
         
    
    
   
     
    
      
     
       
        
       
   
   
   
   
    
  
  
   
   
   
KE! 
ABS 
This 
pixe 
The 
proc 
whil 
For 
data 
papt 
An 
segr 
obje 
cate 
thre 
leve 
regi 
met 
dete 
imp 
dim 
anal 
met 
sevi 
sam 
Sin 
witl 
dev 
con 
upo 
ord 
and 
des 
Ap: 
bee 
ima 
(19 
edg 
thai 
con 
den 
con 
mo. 
larg