Description
Accurate characterization of blob-like filaments in the scrape-off layer (SOL) of magnetically confined plasmas is essential for understanding cross-field transport and refining stochastic models of turbulence in fusion devices. Velocity estimation from imaging diagnostics such as gas-puff imaging (GPI) and beam emission spectroscopy (BES) is often hindered by the barber pole effect, where elongated and tilted structures moving at an angle to their symmetry axis produce erroneous velocity components in time-delay estimation (TDE) methods. This work introduces a robust velocity estimation technique based on two-dimensional conditional averaging (2DCA) that overcomes this limitation and is tailored for the coarse spatial resolution typical of these diagnostics. The method applies 2DCA to characterize the average shape and evolution of high-amplitude coherent structures passing through a reference point. The position is then tracked as the centroid of a contour at a fractional amplitude level, from which the velocity is estimated. We validate the method using synthetic datasets obtained with the blobmodel library, modeling uncorrelated Gaussian pulses with aspect ratio 4 and variable tilt angles relative to propagation direction. Results show accurate recovery of both velocity components across tilt angles, with mean square error significantly lower than TDE methods, which exhibit large barber pole biases. The method proves robust against noise, overlap, resolution variations, and background dynamics. The method is compared with similar tracking methods based on two-dimensional cross-correlations that have widely been used in the literature, showing that the 2DCA-based method yields more robust estimates when background dynamics, such as poloidally propagating waves, severelly bias cross-correlation-based estimates. This approach enables reliable blob velocity statistics for GPI/BES data, enhancing SOL transport predictions and stochastic modeling. The implementation is openly available on GitHub.