Abstract: Since the performance of traditional beamforming degrades when steering vector error and covariance matrix error are contained, an algorithm based on interference and noise matrix reconstruction is proposed to solve this problem in this paper. Firstly, the reconstructed covariance matrix is estimated by discrete summation of the signal and interference in their arrival region, then the steering vector corresponding to the reconstructed matrix is estimated by the eigenvector corresponding to the largest eigenvalue of the reconstructed matrix. The interference and noise matrix are calculated based on the power estimation. The interference power is calculated directly based on spectrum, the noise power is calculated by summing the Capon spectrum outside the desired signal and interference region, then the interference plus noise matrix is reconstructed, so that the optimal weighted vector is obtained. Simulation results show that, the proposed algorithm is well robust to steering vector error and element displacement error, and still works effectively when snapshots are limited.