Résumé : This paper examines data mining approaches to classify the state of road surfaces by simple sensors such as acceleration sensors and gyroscopes. The work aims for classification of road material and the recognition of irregularities such as potholes or railroad crossings. The sensor information therefore is transformed into frequency-based features, which are automatically rated and discussed. The best features are used to design classification routines. All routines are finally implemented into an open-source MATLAB-toolbox visualizing classifier results upon road maps, enabling the user to manually validate the results. All results are discussed using an extensive exemplary data-set.