This paper presents a signal direction-of-arrival (DOA) estimation approach that possesses excellent performance in spatial and temporal correlated signals environments. Firstly, an algebraic solution of the null subspace is derived based on the Vandermonde structured steering vectors of uniform linear array when all the sources possess identical nonzero mean value, and then, the DOA estimation is obtained with a polynomial rooting method. The novel algorithm performs much better than the conventional algorithms in the situation that the sources are closely spaced or correlated, and simulations have verified the validity of the algorithm.