Burrs (excess material) are usually squeezed out around the periphery
of the mould, and are irregular and vary considerably between parts. T
his paper aims at the development of a machine vision system for autom
atic detection of burrs and peripheral defects of casting parts. This
non-contact detection result can be applied to automatic deburring sys
tems and used for automatic inspection of peripheral breakdown. This r
esearch concentrates on flat casting parts comprising piecewise smooth
curves on the boundaries. The detection procedure is based on the fac
t that the irregular burrs show high curvature changes in a small segm
ent of the boundary, whereas a smooth segment without burrs presents a
succession of low curvature points. Given a casting part with burrs,
we arbitrarily select a few boundary points from the smooth segments a
nd connect these points to form a polygon. The matching process then f
inds a corresponding polygon from the model part. The transformation p
arameters of rotational angle and translation between the part in the
image and the model part are evaluated from these two equivalent polyg
ons. The proposed matching algorithm is computationally fast and the t
ime complexity is bounded by O(m), where m is the number of points on
the digital boundary of the model part.