Due to the large terrain fluctuations, dense vegetation and some buildings for hydropower projects in high mountain areas, the problem of low accuracy of airborne LiDAR point cloud filtering always exists. A point cloud filtering method that includes the point cloud echo characteristic, a progressive triangulated irregular network densification(PTD)algorithm, an improved surface fitting algorithm and a refined ground point cloud filtering method was proposed. Based on the denoising of the airborne LiDAR point cloud data, the echo characteristics of the vegetation point cloud is first used to remove some vegetation points, and the PTD algorithm is applied to perform two iterative calculations to obtain a partial ground point set which is used as the seed point of the improved surface fitting algorithm to perform surface fitting of the grid area to obtain the ground points in the original point cloud data. Finally, the point cloud is refined to remove the low-level vegetation points included in the ground points to get the final ground points set. The measured data of the DJI-M600 equipped with the HS-600 airborne LiDAR measurement system was selected for experiments, and the proposed filtering method was compared with the PTD algorithm, surface fitting filtering algorithm, area growth filtering algorithm and morphological filtering algorithm. The results show that compared with the other four methods, the maximum reductions of the first type error, the second type error, and the total error are 6. 25%, 1. 82%, and 2. 93%, respectively. The proposed method is more suitable for the filtering process of airborne LiDAR point cloud data of hydropower projects in high mountainous areas.