In this paper, we present an improved line-of-sight guidance law (LOS) for path following of underactuated unmanned surface vehicles. In the proposed approach, the sideslip angle is treated as an unknown system state that is estimated simultaneously with the cross-track error using an augmented extended Kalman filter (AEKF). Simulation results demonstrate that the proposed guidance law exhibits faster convergence and better path following performance compared to the available LOS methods.