Motivated by the performance of the shape descriptor based on the Largest Intersection and Projection (LIP) signature, we propose a three-dimensional extension in link to concrete application. By projecting 3D shape onto their principal planes and analyzing the resulting profiles, we extract compact and interpretable geometric features. Descriptors obtained from these features are then used for 3D object classification. The proposed method is evaluated on both the ModelNet benchmark and a real-world dataset of tree bark defects. Results show that the descriptor effectively captures shape variations while enabling a significant reduction in feature dimensionality.