We propose a workflow to obtain the path-average dispersion curves, which are the input of SW tomography, from 3D ambient noise records. The workflow starts with a pre-processing step, which separates the time windows at which significant surface-wave energy has been recorded and sorts them, based on the azimuthal direction of the SW origin. For each direction, aligned pairs of receivers are found and the path-average dispersion curves are extracted for each pair. The availability of long records allows stacking, improving the data quality. We applied the proposed workflow on a dataset recorded in 2018 in the Siilinjärvi mining site in Finland, with the purpose of increasing the knowledge of the extension of the phosphate mineralization. Comparison with active dispersion curves proves the reliability of our results and indicates that the passive data allow deeper investigation, increasing the possibility of mapping deeper mineralization targets.