3D Object Classification using 3D Racah Moments Convolutional Neural Networks

In this paper, we propose a new architecture of deep neural network called 3D Racah Moments Convolutional Neural Network (3D RMCNN) to improve the classification accuracy and reduce the computational complexity of a 3D pattern recognition system. The proposed architecture consists of fusioning the concepts of image Racah moments and convolutional neural network (CNN), largely utilized in pattern recognition applications. Indeed, the advantages of the moments concerning their global information coding mechanism even in lower orders, along with the high effectiveness of the CNN, are combined to make up the proposed robust network. The aim of this work is to investigate the classification capabilities of 3D RMCNN on 3D shape datasets. The experiment simulations with 3D RMCNN have been performed on SHREC 2011 and ModelNet10 databases. The obtained results show high performance in the classification accuracy of the proposed model and its ability to decrease the computational cost by training low number of features generated by the first moment layer.