In-plane permeability of textile fabrics is often characterized by one-dimensional rectilinear flow experiments, displaying limitations related to potential race-tracking effects and the need to run experiments at several fiber volume fractions. We propose a practical approach to alleviate these drawbacks. Resin flow front location is detected by image processing of videos and coupled to a Control Volume Finite Element solver with an error minimization routine based on Levenberg-Marquardt method comparing numerical and experimental fill times of all control volumes. Permeability of an isotropic random mat was characterized through reference experiments: cases with intentionally introduced race-tracking and cases with varying permeability along the resin flow direction with three different sections and fiber volume fractions. The method led to an efficient permeability determination and provided valuable insights on the statistics of spatial permeability distribution while highlighting the benefits of a thresholding algorithm for interpretation of permeability experiments in the presence of race-tracking.