When employing spectrum detection and interference countermeasures against non-cooperative unmanned aerial vehicles (UAVs), spectrum interference devices inevitably generate leakage signals outside of the main lobe direction, significantly deteriorating the DOA estimation accuracy of adjacent spectrum devices. To address this issue, an adaptive virtual array extension CAPON algorithm suitable for interference backgrounds is proposed. Initially, an optimized DOA estimation method of “coarse estimation + fine estimation” is designed, obtaining coarse estimation results through the CAPON algorithm. Subsequently, the current number of array elements is adaptively determined to meet the DOA estimation accuracy requirements based on the steepness of the coarse spatial spectrum peaks, thus setting the scale of array extension to avoid the problem of computational complexity surging due to blind array expansion. Finally, by employing an array translation method for virtual extension of the array antenna, the DOA estimation resolution can be enhanced without altering the actual configuration of the array antenna. Numerical simulation results demonstrate that the proposed method can still accurately estimation the DOAs of UAV image transmission sources under interference backgrounds, with DOA estimation accuracy essentially consistent with that in non-interference environment.