Feature extraction of the surface and form of stockpiled materials is performed for achieving the automation and intelligence of warehousing, and it provides the analysis basis for the automatic storage and acquisition control of the materials. First, the stockpiled material is scanned by using LiDAR to determine its morphology and coverage characteristics, a 3D point cloud is obtained, and a fusion algorithm is used to preprocess the material. Second, the supervoxel clustering of point clouds is performed based on the difference of the surface normal vector and spatial distance. Finally, the convex surface is extracted from the 3D point cloud surface after clustering by using the concave and convex judgment method to analyze the surface shape of the stockpiled material. The experimental results show that the method can precisely recognize the surface characteristics of the stockpiled material, and the recognition error is less than 3.11%. The proposed method can be directly applied to different stockpiled material scenarios without training.