The paper describes a novel technique to perform real-time system identification of a fixed wing Unmanned Aerial Vehicle (UAV) using output-only analysis. The method estimates the dynamic behaviour of the system under its real operative conditions with fast and automatic analysis resorting to Operational Modal Analysis (OMA). The cornerstone of the developed algorithm is the Polyreference Least-Squares Complex Frequency-Domain (PLSCF) estimator which is computationally efficient and presents a clear stabilization diagram. This allows for the automatization to be achieved by means of a multi-stage clustering process. The technique is assessed with data acquired from flight tests which are carried out with a fixed-wing UAV.