Abstract: In order to solve the coherent problem of orthogonally matched pursuit (OMP) algorithm in sparse reconstruction, an improved de-coherent method is proposed by using vectorized autocorrelation matrix. The improved method reconstructs the angle only by the vectorized one-dimensional vector, which can reduce the influence of the noise and realize the decoherence without knowing the number of the signal source. Compared with the classical OMP algorithm, the sparse reconstruction effect is better. Theoretical analysis and simulation results verify the good performance of the algorithm.