Publications du COGIT

# Integration of historical geographic data into current georeferenced frameworks: A user-centred approach

5th International Workshop on Digital Approaches in Cartographic Heritage - feb 2010

The creation and diffusion of geographical information have considerably increased over the last few years. In order to benefit from this information as much as possible, either to carry out better analysis or to study the temporal evolution of georeferenced data, users have expressed a strong need to couple - or to integrate - their data with other data provided both by producers or other users. The need to integrate historical data is particularly expressed. This need can be easily explained by the fact that historical data contains invaluable information which is often unmapped or not represented in current maps or data. Historical data is thus particularly of interest to ecologists (study of forest evolutions, comparison of ground occupation on various dates, study of climate evolution, etc.), archaeologists and historians, and also to research scientists who work in the field of simulation (research of evolution rules based on historical data).

The goal of all these actors is first to digitize historical data, then to integrate it into a recent georeferenced framework and finally to vectorize it in order to get a meaningful result. In this context, our paper focuses on the integration of vectorized historical geographic data into current georeferenced frameworks. A user-centred approach is proposed to take into account both users knowledge and users constraints.

Users are not necessarily familiar with the georeferencing process, notably with spatial transformations (affine transformations, Helmert transformations, transformations based on a gravitating model, triangulation and rubber sheeting, second or higher order polynomial transformations, thin-plate spline method, etc.). The system would therefore propose the "most adapted" transformation in regard to user needs and the most adapted mathematical way to solve the given problem (even if a least squares adjustment is traditionally used to solve the problem). "Most adapted" means here that the transformation has to satisfy different kinds of constraints. In our proposal, the objective is to minimize distortions and to take into account some possible user constraints.

On one hand, distortions have to be quantified to know how a transformation can accurately map all control points. This can be done by computing the Root Mean Square (RMS) error based on the residual errors (a residual error is the distance between the target control point and the associated transformed source point). This indicator gives a good assessment of the consistency and the accuracy of a transformation between the different control points. Nevertheless, even if the RMS error is low, some residual errors can be particularly significant (e.g. due to a misplaced control points). In this case, a couple of control points can be removed to improve the transformation.

On the other hand, users sometimes need to use georeferenced data for other tasks than those which simply consist in overlapping several layers, e.g. to analyse the orientation of geographical features. The transformation has in consequence to minimize as much as possible length, angular or surface distortions. Information would in this case be added to the system, for example "an affine transformation implies that straight lines remain straight, parallel lines remain parallel, rectangles may become parallelograms".

To go further than the classical georeferencing process, we propose to introduce two additional steps. Firstly, to tackle the fact that the georeferencing process is generally based on a global transformation without consideration of local distortions, we propose solutions of local spatial adjustment. Secondly, knowing that historical data is often available by geographical area and that each area is represented by a map sheet, we raise the problem of connections management between different map sheets.

We illustrate our work through the example of a Cassini map georeferencing process, showing in particular that several georeferencing methods are possible according to users needs.

The goal of all these actors is first to digitize historical data, then to integrate it into a recent georeferenced framework and finally to vectorize it in order to get a meaningful result. In this context, our paper focuses on the integration of vectorized historical geographic data into current georeferenced frameworks. A user-centred approach is proposed to take into account both users knowledge and users constraints.

Users are not necessarily familiar with the georeferencing process, notably with spatial transformations (affine transformations, Helmert transformations, transformations based on a gravitating model, triangulation and rubber sheeting, second or higher order polynomial transformations, thin-plate spline method, etc.). The system would therefore propose the "most adapted" transformation in regard to user needs and the most adapted mathematical way to solve the given problem (even if a least squares adjustment is traditionally used to solve the problem). "Most adapted" means here that the transformation has to satisfy different kinds of constraints. In our proposal, the objective is to minimize distortions and to take into account some possible user constraints.

On one hand, distortions have to be quantified to know how a transformation can accurately map all control points. This can be done by computing the Root Mean Square (RMS) error based on the residual errors (a residual error is the distance between the target control point and the associated transformed source point). This indicator gives a good assessment of the consistency and the accuracy of a transformation between the different control points. Nevertheless, even if the RMS error is low, some residual errors can be particularly significant (e.g. due to a misplaced control points). In this case, a couple of control points can be removed to improve the transformation.

On the other hand, users sometimes need to use georeferenced data for other tasks than those which simply consist in overlapping several layers, e.g. to analyse the orientation of geographical features. The transformation has in consequence to minimize as much as possible length, angular or surface distortions. Information would in this case be added to the system, for example "an affine transformation implies that straight lines remain straight, parallel lines remain parallel, rectangles may become parallelograms".

To go further than the classical georeferencing process, we propose to introduce two additional steps. Firstly, to tackle the fact that the georeferencing process is generally based on a global transformation without consideration of local distortions, we propose solutions of local spatial adjustment. Secondly, knowing that historical data is often available by geographical area and that each area is represented by a map sheet, we raise the problem of connections management between different map sheets.

We illustrate our work through the example of a Cassini map georeferencing process, showing in particular that several georeferencing methods are possible according to users needs.

## Références BibTex

@InProceedings\{Grosso10, author = "Grosso, Eric", title = "Integration of historical geographic data into current georeferenced frameworks: A user-centred approach", booktitle = "5th International Workshop on Digital Approaches in Cartographic Heritage", month = "feb", year = "2010", address = "Vienna (Austria)", }