The recent approach, proposed by Guzman-Inigo et al. \cite{GuzmanInigo2014}, using System Identification to derive a Reduced Order Model from snapshots of a flow is applied to a transitional boundary layer growing over a flat-plate. It is shown that such an approach can indeed be applied to experimental PIV snapshots. Using a proper learning dataset and a proper local sensor, it is shown that the evolution of boundary layer can be properly estimated from the time evolution of the local probe and with no more than ten POD modes for the Reduced Order Model. The influence of the various parameters on the efficiency of the system identification technique is discussed.