An approach to Haniwa face similarity evaluation using 2D images derived from 3D measurement points
Yoshitaka Sasaki, Akio Kimura, Xin Lu, Kouichi Konno · 2024
This paper proposes a quantitative method for calculating and evaluating the facial similarity of human Haniwa based on 2D images generated from 3D point clouds measured from actual Haniwa artifacts. In the field of archaeological research in Japan, it is extremely important to clarify the excavation information, production process, and artistic value of human Haniwa, and to classify (group) each object appropriately. In order to achieve this goal in a quantitative way, Lu et al. proposed a quantitative similarity evaluation method to directly use the 3D measurement point cloud of Haniwa. They reported that the result is to some extent consistent with the clustering result obtained subjectively by archaeologists in the past. However, the computational calculation time required for 3D point matching with repetitive algorithms remains as a problem. Therefore, this paper proposes another simpler method for objective and quantitative similarity evaluation to consume less computational resources by appropriately converting 3D point clouds into 2D images. The paper also shows that the clustering result is as good as the previous research.