Subspace-based algorithms, which provide high spectrum resolution and is of less time-consuming, are the classical and commonly used methods. However, there are still problems when applying the algorithms to MIMO radar. To solve the problem of the DOA estimation of multiple targets for MIMO radar, a new method based on improved spatial smoothing is proposed. The concept of "equivalent sources" is introduced, based on it, the reason why several subspace-based algorithms would fail when the number of targets is more than the number of transmit antennas is analyzed. The mechanism of the proposed method to overcome the limitation of the number of transmit antennas in the condition of sufficient receive antennas is derived and expounded, which makes the MIMO radar estimate more targets in the case of fewer transmit antennas. The simulation results demonstrate that the proposed method has a better estimated performance than the TDS.