The invention discloses a LiDAR point cloud data recovery based on sparse coding. The method comprises that 1) original point cloud data is re-sampled via software ArcGIS, and point cloud data after rasterization is obtained and forms a point cloud data matrix; 2) extraction of a mask layer matrix is carried out on the point cloud data matrix and applied to the point cloud data matrix; 3) deficient data in the point cloud data matrix obtained in the step 2) is repaired via a self-adaptive learning process of sparse coding; and 4) operation is carried out on a dictionary matrix and a sparse vector matrix obtained in the step 3) via weighted averaging, and repair of the deficient data in the point cloud data matrix is completed. The method can be used to overcome cloud point data deficiency caused by relatively complex landform, highly-absorbent ground objects, non-overlapped degree of air strips, shielding by ground objects, and the like; more detailed features can be remained, complete DSM and DEM models are generated, and subsequent processing and application are realized more effectively.