In the direction of arrival (DOA) estimation, when the unit spacing of the receiving antenna array is greater than the half wavelength of the incident signal carrier, the manifold ambiguity will occur, which poses a challenge to distinguish the real signal DOA. Based on a subarray level large spacing uniform dual-polarization array, this article proposes to randomly change the polarization state of a certain unit in each subarray, so that the corresponding steering vectors of different subarrays are independent of each other. It can break the linearity between the steering vectors under a large spacing uniform array and suppress manifold ambiguity. The simulation and experimental results show that this method can effectively remove spurious peaks in the spatial spectrum of the multiple signal classification (MUSIC) algorithm and achieve a high-precision DOA estimation.