Surface-common-feature descriptor of point cloud data for deep learning

This paper addresses the problem of feature extraction for 3d point cloud data by using autoencoder. Deep learning is one of the most active fields of artificial intelligence, especially in a variety of visual applications, such as image classification and object recognition. However it has not been successfully applied on 3d point cloud data. In this paper, a new method of analyzing the point cloud data is proposed. The method aims to convert the point cloud data to a surface-condition-feature map, which is very effective and useful in pre-training by autoencoder. The surface-condition-features in this paper are defined as upward inclined, downward inclined, upward curved, downward curved, edge and flat, where those features are converted from surface normal vectors.