It is always a challenging issue for radar systems to estimate the height of a low-angle target in the multipath propagation environment. The highly deterministic maximum likelihood estimator has a high accuracy, but the errors of the ground reflection coefficient and the reflecting surface height have serious influence on the method. In this paper, a robust es-timation method with less computation burden is proposed based on the compound reflection coefficient multipath model for low-angle targets. The compound reflection coefficient is es-timated from the received data of the array and then a one-di-mension generalized steering vector is constructed to estimate the target height. The algorithm is robust to the reflecting sur-face height error and the ground reflection coefficient error. Fi-nally, the experiment and simulation results demonstrate the validity of the proposed method.