Segmentation Method for Jomon Pottery Patterns Based on 3D Measured Point Clouds

In archaeology, methods that facilitate the observation of surface patterns on artifacts are essential.Methods for unfolding the entire surface of pottery have been proposed, However, when applied to artifacts like Jomon pottery with natural manufacturing irregularities, these methods often emphasize the non-pattern surface irregularities, making it difficult to identify the actual decorations accurately.First, the point cloud is fitted to a quadric surface and sliced into crosssections based on the main axis.Each cross-section's point cloud is then fitted with an ellipse, using outlier removal to ensure accuracy.By evaluating the distance between the ellipses and the point cloud, decorative points are identified.However, due to manufacturing inaccuracies, cross-sections may not always fit well to ellipses, and and it may be difficult to judge the patterns accurately.Therefore, contour analysis is performed on the distances between the point cloud and the ellipse in the cross-sections.Since the points on Jomon pottery do not change smoothly, the distance graph between the ellipse and points is smoothed with a Gaussian kernel, and then analyzed to improve pattern recognition.