Abstract: According to the characteristics of weak target detection algorithms, a new multiple input multiple output (MIMO) radar detector is proposed. Based on the classical linear model, the noise components with unknown statistics are estimated via maximum likelihood (ML) method. With the estimations, the close expression of false alarm rate and detection probability is derived on the conditions of uncorrelated and randomly correlated channel. The detection algorithms of MIMO radar with unknown parameters at random channel are analyzed and evaluated. The target detection with the noise of unknown statistic on MIMO radar is solved and the capability of the proposed method is validated with the simulation results.