(9) 
ns for the ini- 
ed as: 
e and elec- 
o the initial 
iance matrix 
the accuracy 
0s and chart 
gure 4 and 5. 
middle of the 
ey areas are 
ter. The tick 
hese images 
r Mannheim. 
he radar im- 
also plotted. 
hing process 
te significant 
  
  
  
  
  
  
  
Figure 4: Distance vectors between chart and radar 
image 
distance vectors. In figure 5, the same radar image 
is shown. Now the electronic chart is superposed 
to the radar image based on the estimated position 
and heading of the vessel. Here the uncertainty of 
the matching process is visualized by the rectangle 
around the own ship. This rectangle symbolizes the 
30 c - confidence ellipsoid of the positional correc- 
tions. It is evident that the accuracy of matching in 
transversal direction is much higher than in the longi- 
tudinal direction. 
4.2 Matching stationary targets and electronic 
chart 
The matching algorithm described in section 4.1 
provides enough information about the position and 
heading of the ship on most inland waterways. Its 
basic requirement is that the major part of the con- 
tours stored in the electronic chart are also visible 
in the radar image. This is always true for canals 
with almost constant water level. On a free flowing 
stream like the major part of the river Rhine, the water 
level can vary widely. At the lower Rhine area from 
Duisburg towards the North Sea, large parts of the 
river banks are often flooded and therefore no longer 
visible. To allow a safe navigation on that part of 
the river, the banks and the groynes are equipped 
with radar reflectors that remain visible even at a 
69 
  
Figure 5: Accuracy of matching 
high water level. These radar reflectors are stored 
as landmarks in the electronic chart. They are used 
by another matching process that will be described in 
this section. 
A multiple-target tracking algorithm (cf. section 5) 
is implemented within the integrated navigation sys- 
tem. This algorithm searches for radar objects on 
the waterway and tracks them through an image se- 
quence. When a radar object is found that cannot 
be associated with an existing track, a new track is 
initiated. As there has been no other information on 
this track, it is marked as ‘tentative’. If there have 
been measurements for this new track in some sub- 
sequent radar images, the estimates of the position 
and velocity of the tracked object become more accu- 
rate and the track is marked as confirmed’, meaning 
that the track is with a high probability caused by a 
really existing object. 
Based on the estimated velocity and its covariance 
the tracked objects can be classified as stationary or 
moving. For the purpose of matching with the elec- 
tronic chart only confirmed tracks also classified as 
stationary are used. The track coordinates estimated 
by the multiple-target tracking algorithm are used as 
image measurements. In a next step, the distance 
vectors from the image measurements to the near- 
est landmarks of the electronic chart are computed. 
The motion of the own vessel during the acquisition of 
the radar image is already taken into account by the