Vjacheslav Listsyn 
  
contours are performed (we consider matched filter features later). We will be searching for pairs of contours — one dete 
from “positive” part of difference image and other from it’s “negative” part — that touch each other in the bottom point simi 
and have beforehand determined angels of incline difference. In order to increase operational stability of algorithms 
threshold processing and binarization of difference images and suppression of isolated points are performed after The 
contour selection. cont 
calc 
We say that object is detected if output response of the matched filter has the value exceeding some threshold value. In sem 
case when the matched filter gives a multiple response, these multiple responses are combined into connected regions. max 
In each region the maximum is identified. Position of the object and the distance to it could be determined by the pixel sem 
corresponding to this maximum. con 
cont 
Block diagram of the algorithm is presented on Fig.1. 
The 
mis; 
3 MATCHED FILTER DESCRIPTION algc 
trap 
The matched filter consists of a doubled rectangular mask. One part of the mask is designed to detect positive contours 
inclined in the pre-determined range, and the other part is designed to detect negative contours inclined in the pre- 
4M 
  
Fig: 
Fig. 
usir 
ordi 
witl 
Fig. 
per! 
ima 
  
Fig 
that 
dift 
sub 
vali 
the 
pro 
line 
pat 
       
    
       
  
   
Subtraction « of i Images 
I 
HE ee m 
    
TE 
  
   
  
Fig 
pro 
tria 
ap 
Threshold processing, binarization and | 
suppression of isolated points | 
A mm SETI 
    
   
TTT 
  
Fig 
Inv 
ste 
the 
no 
car 
   
SAE 
  
      
   
    
- Combination of filter responses into connected regions. 
Search for the) maximum por in the g region 
  
  
eure l. 
  
536 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.