Method for locating eye points on objects subject to size variations

Claim 1

A method for locating eye points on lead frame like objects that are subject to size variations, comprising the steps of: acquiring a video camera image of a reference object having a known position in a PRS system, isolating a teach window area in the camera image field which contains a reference object or eye point, generating an edge image of interconnected pixel lines at high contrast edge portions of said reference object, said pixel lines being selected to define unique object shape and eye point, generating a teach eye point in said teach window whose taught location is arbitrarily defined as the center of the teach window, generating a large number of teach vectors between said teach eye point and said edge image of said interconnecting pixel lines, reducing the large number of teach vectors by eliminating some of the large number of vectors to provide a reduced set of teach vectors having a common tail at said teach eye point and heads which are spaced uniformly along said interconnected pixel lines of said unique object shape, storing said teach eye point and said set of teach vectors which define the location of the teach object shape relative to the video camera, acquiring a video camera image of a find object to be located relative to said video camera, generating interconnected pixel lines at high contrast edge portions of said find object, generating sets of find vectors from vector candidate eye points in said image of said find object, comparing said stored teach vectors with said sets of find vectors, scoring the find vectors that compare best with said teach vectors, selecting only the candidate eye points having the highest scores, comparing the selected candidate eye points with the teach eye point, and selecting the one candidate eye point which is closest to the teach eye point as the accurate X-Y position eye point of the find object.

Claim 2

A method for locating eye points on lead frame like objects that are subject to size variations, comprising the steps of: acquiring a video camera image of a reference object having a known position in a PRS system, isolating a teach window area in the camera image field which contains a reference object or eye point, generating an edge image of interconnected pixel lines at high contrast edge portions of said reference object, said pixel lines being selected to define unique object shape and eye point, generating a teach eye point in said teach window whose taught location is arbitrarily defined as the center of the teach window, generating a large number of teach vectors between said teach eye point and said edge image of said interconnecting pixel lines, reducing the large number of teach vectors by eliminating some of the large number of vectors to provide a reduced set of teach vectors having a common tail at said teach eye point and heads which are spaced uniformly along said interconnected pixel lines of said unique object shape, storing said teach eye point and said set of teach vectors which define the location of the teach object shape relative to the video camera, acquiring a video camera image of a find object to be located relative to said video camera, generating interconnected pixel lines at high contrast edge portions of said find object, generating sets of find vectors from vector candidate eye points in said image of said find object, comparing said stored teach vectors with said sets of find vectors, scoring the find vectors that compare best with said teach vectors, selecting only the candidate eye points having the highest scores, comparing the selected candidate eye points with the teach eye point, and selecting the one candidate eye point which is closest to the teach eye point as the accurate X-Y position eye point of the find object.

Claim 10

A method for locating eye points on lead frame like objects subject to size variation, comprising the steps of: establishing a range of size variations of the objects, storing a pixel map of a reference object which contains a reference eye point in a PRS system, storing a centroid location of said reference object, generating a set of reference vectors which defines the location of said reference object, replacing the reference object with a find object to be located, storing a pixel map of the find object to be located, generating a set of find vectors for each candidate eye point in the pixel map of the find object, comparing each set of find vectors with said reference vectors, scoring each set of find vectors for coincidence with said reference vectors, selecting the candidate eye points having the highest scores for coincidence, generating a centroid of a find object for each selected candidate eye point, comparing the centroids of said find object with the centroid of the reference object, and selecting the candidate eye point of the find object being sought whose centroid most closely matches the centroid of the reference object.

Claim 11

A method for locating eye points on lead frame like objects subject to size variation, comprising the steps of: establishing a range of size variations of the objects, storing a pixel map of a reference object which contains a reference eye point in a PRS system, storing a centroid location of said reference object, generating a set of reference vectors which defines the location of said reference object, replacing the reference object with a find object to be located, storing a pixel map of the find object to be located, generating a set of find vectors for each candidate eye point in the pixel map of the find object, comparing each set of find vectors with said reference vectors, scoring each set of find vectors for coincidence with said reference vectors, selecting the candidate eye points having the highest scores for coincidence, generating a centroid of a find object for each selected candidate eye point, comparing the centroids of said find object with the centroid of the reference object, and selecting the candidate eye point of the find object being sought whose centroid most closely matches the centroid of the reference object.