In order to solve the problem of UAV's autonomous landing when GPS signal is failed, a UAV landing method based on FastSLAM algorithm is adopted. First, based on the ground speed measured by Doppler radar, the motion model is described by nonlinear equation with Gaussian noise. The UAV uses its laser radar to obtain the distance and azimuth relative to environmental landmarks for constructing the system observation model. Path estimation and environmental landmark estimation are implemented by particle filter and extended Kalman filter respectively. The velocity estimation is obtained by using tracking differentiator of the arctangent form. Simulation results show that by choosing appropriate number of particles and environmental landmarks, FastSLAM algorithm can meet the requirement on landing precision and real-time performance, and is an effective landing navigation method.