2.1 Preparing the Map for Matching 
The Ordnance Survey map of the test site is shown in Figure 
1(a) which consists of 900 x 500 pixels in size. It is desired 
to have a map in vector format, but in the present work this 
is available from the Ordnance Survey only in raster format. 
It is essential to convert the map into vector format which 
can be used for matching with the image. 
A good matching can proceed based purely on the knowledge 
of where edges of images and boundaries of map exist along 
with their gradient direction information. The following 
steps are taken to determine map region boundaries of 
buildings and their gradient direction. 
2.1.1 Converting Buildings to Solid Objects 
The Farnborough map area shows names of roads, roads and 
large buildings with stipples. In order to use the map for 
matching it is necessary to remove the stipples, roads and 
names from the map and produce boundaries of the 
buildings. This section discuss a method to remove the 
stipples to convert buildings to solid objects and the next 
two section deal with the removal of roads and names. 
A powerful and flexible software, Human Information 
Processing Software (HIPS) system based in a UNIX 
  
  
  
  
  
  
  
   
  
  
  
  
   
  
  
   
  
  
  
  
   
   
  
  
   
environment is used for processing images. But, at this 
stage the conversion of buildings to solid objects is 
performed manually by using the drawing package in the 
Erdas Image system, because it is not available in the HIPS 
system. 
All the features in the map have a value of digital number 
(DN) 255 on a background value of DN 0. The map is 
transferred to the Erdas Imagine system from HIPS system to 
give one unique DN value (127) to buildings to achieve 
homogeneity and converting buildings as solid objects. The 
solid buildings are separated from other objects like lines or 
letters of names attached to it by simply using the drawing 
package as shown in Figure 1(b). The building solid object 
map is then transferred to HIPS system for further 
processing. 
2.1.2 Unique Identification to all Objects 
It is necessary to remove small objects and clutter from the 
map which are not required for matching. This can be done 
by performing two steps. Firstly by giving unique DN value 
to each object which is described in this section and 
secondly by the size (number of pixels) of objects to be 
threshold. 
  
Figure 1. a) Ordnance Survey 1:10 000 map of Farnborough area, b) Map of buildings as solid objects of 
Farnborough area, c) Map of Farnborough area with unique identification to all objects, d) Map of Farnborough area 
with clutter information removed, e) Map of Farnborough area with building-boundary regions, f) Gradient 
direction of map boundaries of Farnborough area. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
  
  
  
  
Ar 
the 
At 
pi 
fre 
ob 
ch 
ot] 
ob 
ob 
bu 
"Te 
th 
ri 
e 
th 
ir 
tc 
le