Airborne LiDAR point clouds have become important data sources for DEM generation recently; however the problem of low precision and low efficiency in DEM production still exists. This paper proposes a new technical scheme for high-precision DEM production based on airborne LiDAR point clouds systematically. Firstly, an elevation and density analysis method is applied to filter out outliers. Secondly, ground points are detected by an improved filter algorithm based on the hierarchical smoothing method. Finally, feature lines are extracted by the planar surface fitting and intersecting method, and a simple data structure of feature lines preserved DEM is proposed to achieve reconstructing high accuracy DEM, combing feature lines with ground points. Experimental results show that the proposed scheme is able to compensate for deficiencies of existing DEM reconstruction techniques and can meet the needs of high precision DEM production based on LiDAR data.