Abstract: When applied to the canopy height inversion, the spaceborne laser altimetry system cannot accurately extract the ground position from the echo signal due to factors such as large spot size and complex surface. Aiming at this problem, a new ground return position extraction and vegetation canopy height inversion algorithm was designed to classify the ground waveform and calculate the weighted elevation of effective candidate peak. The geoscience laser altimeter system (GLAS) echo data in the northwestern region of Vermont was processed by this method,so as to realize the ground elevation extraction and canopy height inversion. And the processing results were verified by the Lidar measurement data during the same period, as well as compared with the inversion results of Gaussian decomposition method. The results show that the ground return position extracted by the proposed method is very close to the distribution of ground point cloud collected by the airborne Lidar, and the ground extraction accuracy is better than that of the Gaussian decomposition method. The root-mean-square error (RMSE)and correlation coefficient of GLAS canopy height inversion result are respectively 2.82 m and 0.81, which are in good consistency with those of the airborne laser Lidar and superior to those of the Gaussian decomposition.