To deal with the effect of the nonlinear system error on the terahertz radar imaging, a system error correction algorithm based on minimum entropy is proposed. In the real terahertz inverse synthetic aperture radar imaging experiments, the intrinsic system error influences the echo phase and disperses the range profile, which will lead to the image degradation. After theoretical analysis of the error formulation, the phase error compensation model for 1D range profile is established. It is found that this error can be skillfully removed by the minimum entropy metric. Experimental results show a higher efficiency and better adaptability of the proposed method compared to that based on the reference reflection.