ABSTRACT

Because of their importance as historical sources, old maps are steadily becoming more interesting to researchers and public users. However, the users are no longer satisfied only by simple digitization and online publication. Users primarily require advanced web tools for more sophisticated work with old maps.

This paper is concerned with classification of digitized old maps in form of raster images. An automatic classification of digital maps is useful tools. This process allows to automatically detect areas with common characteristic, i.e. forests, water surfaces, buildings etc. Technically it is a problem of assigning the image’s pixels to one of several classes defined in advance. If the map is georeferenced the classified image can be used to determine the surface areas of the classified regions, or otherwise evaluate their position.

Unfortunately quite substantial difficulties can be expected when attempting to apply these tools. The main cause of these difficulties is varied quality of digitized maps resulting from damage caused to the original maps by time or storage conditions and from varying scanning procedures. Even individual maps from the same map series can differ quite a lot.

The review of the main classification methods with special emphasis on the Bayesian methods of classification is given. An example of this classification and its use is also given. Web application of raster image classification is introduced as well. The web application can classify both individual images and raster data provided via Web Map Services (WMS) with respect to OGC standards (Open Geospatial Consortium). After gathering the data, classification is applied to distinguish separate regions in the image. User can choose between several classification methods and adjust pertinent parameters. Furthermore, several subsequent basic analytical tools are offered. The classification results and registration parameters can be saved for further use.