MACHINE LEARNING METHODS FOR UNESCO CHINESE HERITAGE: COMPLEXITY AND COMPARISONS

Abstract

Recent years have seen the investigation and 3D documentation of architectural heritage becoming more accessible. The digitalization process could be more efficient when artificial intelligence is used in processing point cloud models. This article investigates the use of machine learning classification algorithms and a Multi-Level Multi-Resolution (MLMR) methodology to classify two point cloud projects in China, Nanchan Ssu, and Fokuang Ssu. Performances of multiple algorithms and solutions are compared, proving the applicability of MLMR on the point clouds. The practices pointed out the significance of corresponding features to classification rules and a sound logic in designing a systematic label tree with hierarchical semantic meanings.