COMPUTER ASSISTED PRODUCTION OF PLANIMETRIC DATA : AN INDUSTRIAL APPROACH 
P. NONIN 
ISPRS - Commission IV 
ISTAR - Espace Beethoven - Route des Lucioles 
BP 037 - 06901 SOPHIA ANTIPOLIS - FRANCE 
PURPOSE: 
Space imagery is a cost effective data source for small scale mapping. However, the raster data structure has 
to be converted into a more suitable vector structure for geographic information processing. 
A semi-automated vectorizing system, running on a standard workstation, is described. Up to date image 
processing techniques (spectral/textural analysis, snakes) are employed to speed up planimetric extraction. 
KEY WORDS: Cartography, Classification, Planimetry, Vectorization, SPOT. 
1. INTRODUCTION 
The availability of high resolution remote sensing 
satellites, namely LANDSAT and SPOT, and the set 
up of efficient data delivery facilities, have enabled 
us to acquire and update large image data bases 
from any part of the earth. 
More over, the geometric properties of the CCD 
cameras shipped in those satellites allow 
cartographic data to be produced. 
ISTAR, as several other companies in the world, has 
produced more than 200 geocoded images and 
digital terrain models computed from stereo couples 
of SPOT images since 1988. 
On the other hand, the lowering prices of personal 
computers and workstations encourage digital 
information processing in any domain of activity and 
particularly in geographic applications. 
Unfortunately, most of Geographic Information 
Systems commercially available, cannot handle 
effectively such unstructured data as raster images. 
Computer systems are not qualified, as humans are, 
to analyse the semantic content of satellite images. 
The best way to manage efficiently geographic 
entities is to convert raster images in highly 
structured sets of vector objects describing the 
boundaries of the land cover map and the network of 
linear features. 
Traditional methods require a large amount of human 
interaction for the extraction of both altimetric 
(contour lines) and planimetric features. They also 
require expensive photogrammetric workstations. 
Our purpose was the production of vectorized 
planimetric features from remote sensing images for 
geographic information systems users. 
The data should be cheaper than 
cartographic products and above all, 
updated. 
A computer assisted vectorizing system, running on 
standard UNIX workstations, has been designed. 
A single operator can process an entire SPOT scene 
(60 km x 60 km) within two weeks. 
The data produced describe the geometry and the 
topology of the main planimetric features extracted 
from the raster image (urban areas, forests, 
transportation network...). 
They do not describe “external” information (i.e. 
information not extracted from the images, ex: 
administrative boundaries or demographic data). 
traditional 
recently 
A great demand for such cartographic data can be 
342 
found in several domains like flight simulation, 
transportation planning, telecommunications, 
developing countries cartography. 
2. PLANIMETRIC FEATURES 
In order to reduce production (and commercial) 
costs, a standard general purpose cartographic 
product is proposed. It is a collection of vector 
elements with a layered structure. Each layer 
describe a particular type of features. 
All possible types are listed below: 
built-up [suburban, industrial] area 
transport [road, railway, airport, navigation] facilities 
wood 
waste 
bare soil [rock, beach, sand dune, quarry] 
water [sea, flooded area, mash] 
snow 
This list intends to meet the requirements of general 
rather than specialized needs. It is composed of two 
categories: linear layers (or networks) and area 
layers. 
Each entity within a linear layer is defined as a list of 
bi-dimensional or tri-dimensional coordinates. 
All areas are described by a single topological graph: 
An area entity is defined as a sorted list of arcs and 
nodes. 
  
initial SPOT orthoimage