The present project proposed the development of a dynamic surveying and data analysis approach which would be able to detect the position, nature and degradation of different track defects along the layout. For this, accelerations, both vertical and lateral, were recorded on board the trains. The aim of this process was that it happens in a fully automated way. For this purpose, digital signal processing tools, such as the spectrograms, were used. Such tools decompose the signal and allow its analysis in the time and frequency domain simultaneously. In such representation, track defects showed specific patterns, thus allowing for their identification and classification. These patterns were automatically recognised by means of digital image processing techniques. For this pattern recognition, a training-testing process with the recorded data was carried out.
The output data of the proposed approach could serve as the input data for today-in-use track deterioration models and track maintenance management tools. In the same way, if a first diagnostic is given on the track conditions, duration of visual inspection and later maintenance tasks may be dramatically reduced, thus lowering the respective costs.