An object oriented approach to automatic classification of archaeological features in magnetic prospection data


Magnetometer prospection is commonly used in archaeology for the non–invasive detection, mapping and investigation of buried prehistoric sites. The recorded data can contain numerous anomalies caused by archaeological structures in the ground. State–of–the–art geomagnetic data processing results in geo–referenced maps that traditionally are interpreted within Geographical Information Systems. With the increasing size of surveyed areas, the manual outlining and classification of magnetic anomalies becomes a highly time consuming process. Possibilities for automated classification of the magnetic prospection data prior to the actual archaeological interpretation would considerably enhance the productivity of the archaeological interpretation process regarding magnetic prospection data. Object oriented image processing methods known from remote sensing applications offer a large spectrum of readily available procedures for the automatic and semi–automatic analysis of raster data sets. Suitable algorithms have been adapted and utilized in order to exemplarily analyse a magnetic archaeological prospection data set with the goal of automatically mapping magnetic features, facilitating further archaeological interpretation. This article shows that the presented semi–automated classification procedure is able to replace the manual drawing of anomalies based on the expert knowledge and experience of the interpreter to a large degree.

Near Surface Geophysics