You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Fusion of sensor data, knowledge sources and algorithms for extraction and classification of topographic objects
Baltsavias, Emmanuel P.

International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999
Ph.Gamesson 1 , D.Bruckert 2
1 ACRI, 260 route du Pin Montard, 06904 Sophia Antipolis, France, pg@acri.fr
2 WEUSC, Aparto de Correos 511, 28850 Torrejon de Ardoz - Madrid, Spain, bru@weusc.es
KEYWORDS: Satellite Images, Combination, Feature Matching, Precise Geo-Location.
The main requirement for satellite and airborne imagery exploitation is to make conform (overlay) these images (optical, radar,
infrared, etc) to a working geographic model that can be either a map or another reference image. The GEORIS tool method is to
accurately overlay an image with any other image by using only the image contents. The image analyst community dealing with
operational crisis management and fast decision making has required this important operational constraint. Indeed, in operational
near-real-time context, different images and maps shall be often combined without knowledge of the sensor characteristics. To solve
the current polynomial-registration-method drawbacks, a new state-of-the-art technique has been designed and developed using a
registration model which integrates ground control and tie points, as well as recognised linear features (roads, power lines, railways,
pipe and water networks, bridges, buildings, shorelines, etc). Firstly, linear features are used to calculate the polynomial model
accuracy and estimate its residual quality. Secondly, the linear features are used to define a graph model including thousands of
control points. Then, a numerical process to match the graphs is used to optimally estimate the best registration model. For
mountainous areas or inaccurate maps, a polynomial model cannot be applied, even if it has a high order. The developed GEORIS
method proposes a multiple local fitting model approach. Such a technique has the advantage of taking into account local relief and
increases significantly the matching precision. To improve further the relative pixel positioning accuracy, the last GEORIS version is
able to use a digital elevation model when available.
To exploit a set of digital spacebome and airborne images and
maps either for image analyses and data fusion or more
sophisticated data post-processing like with GIS facilities, it is
necessary to register all of them in a common geographic co
ordinate system.
From an operational point of view, the obtained registration
accuracy conditions the reliability of the derived information,
extracted from the data. Therefore, this step is important and
requires careful methodology to perform the multiple source
data registration (spatial and spectral) and its quality control.
The quality control of the registration accuracy provides to the
user confidence that the set of images and maps can be correctly
matched to extract reliable information.
Classically, geometric registration is based on two different
approaches. The first approach, is based on the use of sensor
geometric models, including observation features and orbital
parameters (Crawford et al., 1996, de Sève et al., 1995, El-
Manadili and Novak, 1996). In this case, requirements are the
availability of complementary information such as satellite
ephemeris, correct sensor models and sometimes digital
elevation models which are usually not supplied with the set of
images (Palà and Pons, 1995). In addition, in a real-time
operation context, such as for natural disasters (floods,
earthquakes, fires, etc), these required additional data are most
of the time not immediately available or even not existing.
The second, more empirical, approach relies on matching
ground control points (Jensen, 1996). The drawback is the
difficulty to automate the process of registration which is done
under the control of an operator, except in some cases where a
database of reference GCPs has been pre-computed (Holm et
al„ 1997).
In this study, the second approach has been investigated. To
satisfy operational needs, the provided images shall be exploited
as they are. In other terms, only spacebome and airborne image
contents are used and contribute with other data sources to
characterise the risk and its temporal development. Therefore, to
support such operational requirements, a procedure and
methodology shall be designed to integrate images, maps and
site reports as they are and produce an accurate overlay of all
this data.
In this context, a system called GEORIS has been designed and
developed by the ACRI company under a WEUSC contract
(Western European Union Satellite Centre, Madrid).
The aim of this article is to provide an overview of the GEORIS
system, and especially of the model used.
The core of the software computes mathematical models to
geometrically register an image with a second image or map
which is used as the reference (the reference image is usually a
geo-coded image called master).