Urban remote sensing investigations.

This paper deals with the research activities achieved by the team TELSAT/06-TELSAT/11/06-TELSAT/T3/D03 of the University of Liege, in the framework of the National research programme on satellite remote sensing (National Scientific Policy Office). The team specialized in urban remote sensing and especially in applications relevant to urban, land and country planning and the monitoring of enevironment. Besides a theoretical approach of the methods of remote sensing, those trends imply a good skill of hybrid geographical information systems and an expertise in satellite mapping.
One the one hand satellite images are radiometric value matrices and must be exploited in a numeric way, on the other hand the needed references to ground-truth require to use visual interpretation methods. As most of the image processing systems have not or had not the necessary functions for extracting and enhancing information, the team has developed or implemented such tools. The main fields of these tools are: mathematical morphology, image segmentation, rasterization and vectorization, contextual filters, transformation of discrete data to spatial continous models. Various methods of optimization of classification are presented, e.g. use of neo-channels (textural or others), zone classification in¬stead of pixels ones, morphological filtering applied to post-classification.
The superimposition of vectorial information produced by SIG on raster classification and the use of map publishing software allow to produce regular satellite maps. The content of landuse maps fulfills some needs of survey for country planning. For some applications, remote sensing methodology is more suited than traditional ones. This is the case to spin up of country into homogeneous regions or for finding net density by use of dasymetric methodology. Another interesting way of working consists in the modelling of landuse map content with a quantitative geographical approach. The raster and numerical nature of satellite images makes easier the construction of continuous models which can be assimilated to statistical surfaces. These surfaces allow analysis of geographical, hierarchi¬cal and chronological aspects of distribution. A specific interest can be found in the comparison of the urban potential statistical surfaces with the central place theory (W. Christaller) and with the economic polarization theory (F. Perroux).
Ecological zoning can be found by coloured composition of 3 statistical surfaces. The 3 most important categories of landuse (water, vegetation and concrete) can be seen as the main structure of biotops and lead to the interpretation in ecosystemic terms. The choice of satellite image allows to cover the full range of scales from cities to regional analysis. The applications concern mainly the "Euregio Mcuse-Rhine" cross-border region and its urbanized core (Liege, Aachen, Maastrichr, Hasselt, etc.). The metho¬dology has also been used outside of the Euregio (agglomeration of Luxembourg, metropolitan area of Lisbon, Warsaw, etc.).

Anyway, spatial landuse maps must be seen as a specific layer of a regional GIS. The Crossing of satellite data with data stemming from other sources allows to set out and to solve a lot of problems relevant to country planning practise.
This paper analyses successively the constraints of applicability, the stade of the art in urban remote sensing, the used methodologies, the completed end- and sub-products, the assessment of the research and an exhaustive bibliography.