E. Buard, A. Ruas




In the context of maps on demand, we wish to create maps according to users needs. Keeping in mind that a map is a creation coming from choices, in particular for data representation and symbolisation, legends have to be cartographically correct to ensure a good visualisation and, as a consequence, a better understanding and efficiency. It means that users’ legends have to be analysed and improved in term of cartographic correctness and semiology. Elisabeth Chesneau at COGIT Laboratory [Chesneau, 2006] has developed automatic methods to analyse and improve the colour contrasts of the symbolised objects in the legend. It does not analyse the contrasts directly in the legend [Brewer, 1997] but locally at the object scale. Her model is based on colour contrasts estimations. In order to obtain coherent results, these estimations have to be reliable. However when it is come to colours, we need cartographic experts to evaluate the contrasts and to give them a relevant score. Colours in cartography contain much knowledge and subjective opinion.


This paper presents our proposal to undertake a tests protocol for colour contrasts evaluation to give to cartographic experts. The work is implemented on Maacol [Dadou, 2005], a module developed at the COGIT Laboratory, helping knowledge acquisition, whatever the knowledge is, and then analysing it using supervised machine learning. For our evaluation, we apply Maacol to cartographic knowledge. First tests have been experimented in choosing specific colours [Jolivet, 2006] representing the most used colours in risks maps. 20 cartographic experts have evaluated 50 test samples in comparing two neighbouring colours. However the scores given to the contrasts were very spread due to a lack of test specifications. That is why we design an accurate test protocol for colour contrasts. It describes the type and the range of wanted answers, the relative position of both colours to analyse, the good visualisation environment for the tests, the tests explanations and the consultation of visual examples of a minimum and a maximum score of contrasts.

A second ongoing test is performed first to validate the test protocol and then to extend the test to a larger range of colours.



Chesneau, 2006: Model for the automatic improvement of colour contrasts in cartography- Application to risks maps, PhD thesis, 372 p.


Brewer, 1997: Evaluation of a model for predicting simultaneous contrast on color maps, professional geographer, 49 (3), pp 280-294


Dadou, 2005: Helping the capture and analysis of expert knowledge to support generalisation, 8th ICA workshop on generalisation and multiple representation, A Coruna, 9p.


Jolivet, 2006: Analyse des contrastes de couleurs – Mise en place d’un test sous le logiciel Lamps2, Master carthagéo, ENSG, rapport de stage réalisé au COGIT, 35 p.