approximate edge direction of the patch (Figure 3). The 
only difference between the pre-rotated template and the 
“original” is the rotation angle. The approximate edge 
direction is derived from the maximum edge gradient 
direction, which in turn is computed from the Sobel 
gradients in a 3 x 3-pixel window. 
Essentially LSM and MPGC are area-based matching 
techniques. For high accuracy edge matching the method 
is transformed into a combination of an area-based and 
feature-based technique. This is achieved by introducing 
as reference template a synthetic (or real) edge pattern, 
which is to be matched with the actual image edges. 
Compared to the conventional feature-based matching 
techniques our method does not require the extraction of 
image edges, but matching is done directly by using the 
original grey value edges. 
2.3 Edge tracking 
The edge measurement procedure has been extended to 
an edge tracking technique, which automatically tracks 
and measures the full edge. The new approximate match 
point for the next patch of the first image is determined 
by using the previously matched position, its local edge 
direction and a user-defined incremental distance (a 
certain number of pixels). In the other images the 
previously determined position is used for the new start 
position. The edge tracking stops either after the 
measurement of a user-specified number of edge points 
or if matching fails, e.g. because of the end of an edge is 
reached. Other termination conditions can be formulated. 
The result of tracking is a polyline which approximates 
the full edge. An additional advantage is the automatic 
determination of good initial values through which the 
template patch 1 patch 2 
  
  
    
number of iterations to determine the unknown 
parameters can be decreased. The tracking is basically 
done in image space. Through the simultaneous 
computation of consistent object space coordinates this 
generates implicitly an object tracking procedure. 
3. IMPLEMENTATION OF THE EDGE 
MATCHING ALGORITHM 
The previously described algorithm is implemented as a 
module in DEDIP (Development Environment for 
DIgital Photogrammetry, Gruen, Beyer 1991). The 
following functional features and options are realised in 
the program: 
e Interactive measurement of image coordinates in 
digital images. 
e Visualisation of the edge matching on the display 
(Figure 4). 
e Interface to bundle block adjustment program in 
DEDIP. 
e Single-point edge matching and edge tracking with an 
unlimited number of images. 
e Edge matching with and without collinearity 
constraints. 
A stand-alone version of the program is also available 
which allows a higher execution speed, since the time- 
consuming visualisation is not used. 
Figure 4 shows the visualisation of the edge matching. In 
the top row are the corresponding image regions around 
the matching point with the start and final positions and 
the epipolar line from thc first image. The frames and 
crosses show the size and the position of the template and 
patch 4 
patch 3 other patches 
TEIIIGIIATIIIIE 
N 
  
ARENA 
NAS3S331»321*1*1** 
  
  
44 
PIII III Isr ss ss sess 
  
   
  
edge 
constraint 
  
  
pre-rotation 
  
template 1 
template 2 
1 
collinearity constraints 
grey value matching equations 
  
Figure 3 Edge matching as implemented 
wua................ 
  
FISTS I IIES. 
  
LCCC CCRC CRN 
URS 
  
A 
  
N 
CTT III I IER 
template 3 template 4