Earthenware is a kind of relic. By analyzing the earthenware, it is possible to conjecture about human activities at the time. Unfortunately, most of them have been damaged when excavated from the ruins. The traditional restoration method for earthenware is typically by manual work. It takes time-consuming and requires special knowledge in archaeology. Since there is always a mass of fragments mixed together color, pattern, and shape of fragments would be mainly used as clues. If entities are difficult to be restored, virtual objects restoration on the computer offers another possibility. In addition, virtual earthenware exhibits made by 3D models and replica earthenware exhibits made by 3D printers have educational values in historical museums and schools. In the present, in order to repair various of relics including earthenware, methods that assembly of earthenware pieces using computer-assisted have been well developed. According to the conventional method, adjacent fragments can be defined by the contour of the fragments and the cross-sectional shape of the rupture plane. Unfortunately, earthenware fragments are much different from other kinds of relic. Its curved surface and thin body make it difficult to calculate a 3D spatial posture. In our study, we proposed a spatial posture optimization algorithm for cylindrical pottery in the Jomon period which based on 3D measurement point cloud. This algorithm can meet all the adjacent fragments information, simultaneously. We have used this algorithm to repair several earthenware fragments successfully and effectively. It turns out to be an approach that deserves widespread use.

View full abstract

Yarn objects are common materials for generating clothes which are essential for creating realistic scenes in computer graphics. Recent methods for generating yarn can mainly be classified into two categories: volume-based and fiber-based. These methods dramatically improved the representation of yarn details and reduced the cost, while the generation process for flyaway fibers still remains a challenge. In particular, some materials having microfibers around their surfaces, such as pile fabric and chenille yarn, are widely used for necessities including towels, car sheet upholstery, and clothes, but the generation processes of these materials are very difficult within the current frameworks. In addition, the previous approach focused only on automating the creation of a single yarn, and as far as the authors know, there exist no methods for producing fabrics from the generated yarn. In order to address these problems, this paper presents a method for generating yarn with additional microfibers and proposes a framework for automatically generating fabrics using our piled yarn with an arbitrary size.

View full abstract

Excavated stone tools peeled from the same mother rock can be restored by assembling stone flakes. In general, excavated flakes are joined manually according to their colors, designs and shapes by trial and error. This process makes joining of flakes more complicated because it takes a long time and needs much effort with special technical knowledge. Therefore, the method is required to substantially reduce the burdens for efficient joining of flakes. Methods to extract flake surfaces of assembling stone tools have been proposed. In such methods, point clouds are separated into flake surfaces using surface boundary and scale drawing. Scale drawing are used to write research excavation reports. Since scale drawing are used to write the reports, abstraction of flake surfaces is very difficult if there are no scale drawing. In addition, the more the number of flakes will be, the more the effort and time will be required for scale drawing. This study examines the method to recognize flake surfaces for not using scale drawing. In our method, the density of point cloud is lowered and ridgelines are extracted from the calculated features by using curvatures. Then, a kd-Tree is built from the candidate points and the point density in each region is obtained. High-dense points are tracked counterclockwise. Thus, a ridgeline is derived from a significant sequence of points. Flake surfaces are recognized by using region growing method from the sequence points. We applied this method to the point cloud data and confirmed an automatic abstraction of flake surfaces.

View full abstract