To retrieve near surface velocity models useful for improving deep imeging surface wave tomography can be used. The advantage of surface wave tomography is that the coherent noise present in the exploration data can be analised to get a high resolution 3D velocity model. The main drowback is related to the picking of the dispersion curves that represent a very time consuming processing step. Automatic picking algorithms are therefore required. In this thesis a contribution to an autopicking algorithm is given with the developmento of a code for detecting outlier dispersion curve and remove them from the dataset automatically. The code has been tested on a large exploration data acquired in the framework of the H2020 Smart Exploration Project