Leveraging the precise near-field (NF) geometry, we present a source localization technique employing multiple cumulant matrices with a COLD array. Systematically, the proposed methodology constructs multiple polarization cumulant matrices, whose number is the same as the total number of physical element, subsequently cascading them into an augmented matrix. Operating eigen-decomposition on this augmented matrix, we extract the signal subspace and achieve effective separation of horizontal and vertical polarization components. The ESPRIT algorithm is then applied to enable simultaneous estimation of angle, range, and polarization parameters. Simulations demonstrate that our method outperforms existing other algorithms in parameter estimation accuracy.