)hgk.mil.tr 
economically 
fogrammetry, 
automatically 
s clearly, the 
‘an computer 
digital aerial 
1ethod which 
photographs, 
re applied to 
n five groups 
ze the roads 
the imagery. 
ethods focus 
1se the other 
exploit them 
ns are left to 
can be seen 
y) 
adiometry) 
ast with the 
y) 
zy) 
(contextual) 
on the other 
ate a road 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
2.2 Road Tracing by Profile Matching and Kalman 
Filtering 
An operator initializes the road tracer by measuring two points 
that indicate a short road segment. Between the two points gray 
value cross sections are computed at intervals of one pixel. The 
model road profile is taken as the average of these cross 
sections. This model profile is used as a template in the profile 
matching. Based on the indicated road segment an initial 
estimate is made of the parameters that describe the road’s 
position and shape. This estimate is used to predict the position 
of the first road profile adjacent to the indicated segment. The 
profile at the predicted position is matched with the model 
profile. The result of this match is a shift between the two 
profiles. This shift is used by the Kalman filter to update the 
parameters that describe the road’s position and shape. In the 
following iterations, the position of the next profile is predicted, 
the profile at this position is matched with the model profile, 
and the road parameters are updated. The road tracer continues 
until some break-off criterion is fulfilled (Vosselman and 
Knecht, 1995). 
In this method least squares profile matching is used to over 
maximizing cross correlation because it can estimate the profile 
shift’s precision which required as input for the Kalman filter. It 
is also possible to model the geometric and radiometric 
transformation between the two profiles by the help of least 
squares matching.. Both of the road position and width can be 
estimated so good results can be obtained whether the road 
width is changing, when cross correlation fails. 
The Kalman filter is a recursive procedure to estimate the 
parameters of a dynamic system and has found many 
applications in navigation (Kalman, 1960; Gelb, 1974). In the 
case of road tracing the parameters to be estimated are the 
parameters that describe the position and shape of the road. 
These parameters are called the state (Vosselman and Knecht, 
1995). 
If the state is not time-dependent, this method does not have a 
dynamic system but if the distance along the road is treated as if 
it were the time variable, then the recursive estimation 
procedure can be used. The Kalman filter consists of two steps: 
— Time update 
— Measurement update 
In the time update an estimate of the state at time /--df is made 
using all observations (i.e. profile matches) that have been made 
up to time /. Thus the time update predicts the state at the future 
epoch /+dt. In the measurement update the results of the profile 
match at time /--dt are combined with the prediction from the 
time update to obtain an optimal estimate for the state at time 
t--dt (Vosselman and Knecht, 1995). 
The profile matching compares the model profile with the road 
profile at the position predicted by the time update. The 
differences between the two profiles are modeled by two 
geometric (shift and width) and two radiometric (brightness and 
contrast) parameters. These parameters are estimated by 
minimizing the square sum of the gray value differences 
between the profiles (Ackermann, 1983). 
After determining the optimal transformation between the 
profiles the matching results are evaluated by three checks: 
519 
— The cross correlation coefficient between the gray 
values of the two profiles after the transformation is 
required to be higher than 0.8. 
— The estimated values of the geometric and radiometric 
parameters should be reasonable. E.g., if the estimated 
contrast parameter has a high value, say 10, the match 
can not be accepted. A contrast value of 10 would mean 
that the gray value contrast in the model profile is 10 
times the contrast in the profile at the predicted 
position. A high contrast value therefore indicates that 
the latter profile hardly contains any signal and most 
likely does not correspond to a part of the road. 
— A match is only accepted if the estimated standard 
deviation of the estimated shift parameter is below 1 
pixel. 
If for one of the above reasons the result of the matching is not 
accepted, the Kalman filter will not perform a measurement 
update but instead continue with another time update. Several 
consecutive rejections of the profile matching can be used as an 
indication for a road junction or the end of the road 
(Vosselman and Knecht, 1995). 
In this method a constant standard deviation of the profile shift 
of 0.3 pixel is used in the Kalman filter instead of obtaining by 
propagating the a posteriori standard deviation of the 
differences between the gray value profiles and based on the 
assumption that the gray value differences have a Gaussian 
distribution. For the Kalman filter processes and algorithms see 
(Vosselman and Knecht, 1995). 
2.3 Semi-automatic Road Extraction Based on Edge and 
Correlation Analyses 
This method works with two feedback loops controlling two 
basic steps and possible interventions of the operator. These 
basic steps are: 
— Extrapolation 
— Extraction 
The inner loop monitors the failures in the extrapolation and 
extraction steps and decides whether the method can proceed 
itself or not. The built-in stopping criterion is based on the 
percentage of the failures in a pre-defined segment of road. 
Three situations may be considered concerning the outer loop. 
First, in the case of successful point extraction the process 
proceeds normally, i.e., a new loop is initialized. Second, the 
process may be automatically finished (e.g., the end of the road 
is detected). Third, the intervention of an operator may be 
required for finishing the extraction process or reentering the 
needed information to restart the process (Poz, 2001). 
In the extrapolation process a parabola used as the road 
trajectory model (McKeown and Denlinger, 1988). The most 
recent points were used to fit the parabola. One characteristic of 
this solution is that only local extracted information is used to 
extrapolate the road trajectory. As a result, some weakness are 
expected whenever the method needs to handle a situation 
involving, for example, an obstacle on a curved road segment. 
To overcome this limitation, a more global solution is proposed; 
involving information located ahead the last extracted (Poz, 
2001).