Table 5 Comparison of axis detection methods
From: Enhancing traditional museum fruition: current state and emerging tendencies
Methods | Exploited property | Pros | Cons | ||
|---|---|---|---|---|---|
Direct methods | Normal based [34] | The normal versor at each node of the surface intersects the axis of symmetry | Quick estimation, requiring no preliminary evaluation | Sensitive to the measurement noise, outliers, surface roughness, to angular spanning and all those factors making the surface not perfectly axially symmetric | |
Iterative methods | Osculating sphere [35] | The centres of the osculating sphere are on the axis of the axially symmetrical surface | The axis estimation does not consider limited singularities and low noise levels | A preliminary axis estimation is required: the quality of the final evaluation depends on the first attempt solution | Sensitive to the measurement noise, outliers, surface roughness, angular spanning and all those factors making the surface not perfectly axially symmetric It fails when the fragment’s shape is close to spherical |
Circle and line fitting ([36]) | Each section of an axially symmetric surface perpendicular to the axis is a circle, whose centre is on the axis | ||||
Profile-based [37] | The 2d profile of the 3d axially symmetric model in cylindrical coordinates z-ρ is a curve | It requires the automatic recognition of axially symmetric point of the profile | |||
Symmetry line-based [29] | The symmetry line of any planar intersection curve of a complete axially symmetrical surface always intersects the surface axis | Only for complete axially symmetric surfaces | |||
THICKNESS versor intersection-based [31] | The minimum wall thickness line always intersects the axis of revolution | The method seems robust in presence of extensive wear, encrustations, and all the typical damage found in a common archaeological sherd | Only for thin walled axially symmetric surfaces | ||