3D Shape Classification Using 3D Discrete Moments and Deep Neural Networks
Zouhir Lakhili, Abdelmajid El Alami, Abderrahim Mesbah, Aissam Berrahou, Hassan Qjidaa · 2019
In this paper, we propose a new model for 3D shape classification based on 3D discrete orthogonal moments and deep neural network (DNN) to enhance the classification accuracy of 3D objects under geometric transformations such as scale and rotation. The proposed model is derived by introducing image moments as an input vector in DNN, frequently utilized in many tasks of pattern recognition. Discrete orthogonal moments have the ability to capture global information from an image in lower orders. The aim of this work is to investigate the robustness of the proposed model to geometric transformations like rotation and scale. The simulations are performed on constructed dataset by applying some geometric transformations on selected objects from the McGill database to evaluate the performance of our proposed model. The obtained results show that the proposed model with Hahn moments achieves high classification rates and robust to geometric transformations than Krawtchouk moments.