natching will be 
ine matching for 
stigates a robust 
elling using high 
an be performed 
ng method using 
w format of the 
he special device. 
NTSC: 525, PAL: 
can be obtained. 
g method using 
m a helicopter at 
gure 1 shows the 
ine extraction and 
the automatic line 
operation with 2 
which called the 
[he height of edge 
rest point, and the 
fluence of noise. 
ig, each both ends 
Jowever, there are 
t image. Then, the 
ration in the only 
ction for the first 
  
Figure 2. Extracted Lines in the First Image 
3.2 Optical Flow Estimation 
In order to perform line matching, each both ends for the 
extracted lines were tracked by optical flow. Although many 
optical flow estimation methods have been proposed, Lucas- 
Kanade method (Chiba, 1998) which has ability of correct and 
fast procedure was adopted in this paper. The optical flow by 
Lucas-Kanade method is calculated by following equation, and 
extracted lines was shown in Figure 3. 
ar VOHEN ES V6) . 
ary er Y 
Hs 
where, 
1(9)= 105.920), (p)= I(r + 81) 
  
3.3 Results of Line Matching 
Detail procedures of the line matching are as follows: firstly, 
line extraction is performed for the first image of the image 
sequences. Similarly, line extraction for the next image is 
performed. Secondly, optical flow is estimated using the first 
image and the next image, and each line positions in the first 
image are moved to amount of the optical flow. Finally, the 
above procedure is successively repeated to the last image, and 
the line matching is achieved. 268 lines for the first image were 
extracted by Canny operation and manual correction was 
performed for removing useless lines, and line matching for 128 
Figure 4. Matched Lines in the Last Image 
(Result of Line Matching) 
lines out of those lines could be performed correctly. Figure 4 
shows the result of the line matching. 
4. EPIPOLAR MATCHING 
The Line matching was performed rather efficiently by the 
above procedures. However, these procedures can not apply for 
all lines due to fragment or multiple. Therefore, the unmatched 
lines were corrected using epipolar line and probabilistic 
relaxation method. The details of the epipolar matching are as 
follows. 
4.1 Epipolar Line 
In order to estimate epipolar line, relative orientation was 
performed by coplanarity condition using the first image and 
the last image. The each both ends for the matched lines were 
used as control points for the image, and the orientation 
parameters (by, bz, c, 9, x) were determined. After the 
orientation, resampling for the last image was performed using 
the orientation parameters. Consequently, epipolar lines were 
estimated. Figure 5 shows the epipolar lines and conjugate 
points. 
  
Figure 5. Epipolar Lines and Conjugate Points 
—501 —