We demonstrate numerical techniques for automatic identification of individual spiral arms in hydrodynamic simulations of astrophysical discs. Building on our earlier work, which used tensor classification to identify regions that were "spiral-like", we can now obtain fits to spirals for individual arm elements. We show this process can even detect spirals in relatively flocculent spiral patterns, but the resulting fits to logarithmic "grand-design" spirals are less robust. Our methods not only permit the estimation of pitch angles, but also direct measurements of the spiral arm width and pattern speed. In principle, our techniques will allow the tracking of material as it passes through an arm. Our demonstration uses smoothed particle hydrodynamics simulations, but we stress that the method is suitable for any finite-element hydrodynamics system. We anticipate our techniques will be essential to studies of star formation in disc galaxies, and attempts to find the origin of recently observed spiral structure in protostellar discs.