Abstract-In conventional statistical STAP algorithms, the existence of interference target in training samples will lead to signal cancellation, resulting in the output SCR falling and the moving target detection performance degrading. The nonhomogeneity detector is an effective way to restrain the outlier, which can improve the covariance matrix estimation by detecting the samples containing outliers and rejecting them, and improve the STAP performance. A new interference target detection algorithm is proposed in this paper, the outlier detection is realized by using the samples' data phase information. Compared with traditional method, the improved algorithm is more sensitive to interfering target with different azimuth and intensity. Simulation results demonstrate the validity of this improved method.