Automated bead detection and detection training

Abstract

A system and method for training a bead detection system can include identifying a starting region on a bead based on a starting indicator. The bead can be analyzed at the starting region to identify bead characteristics. The bead can then be analyzed away from the starting region, based on the identified bead characteristics, to identify a bead profile for use during non-training bead inspection.

1 - 20 . (canceled) 
     
     
         21 . A method for training an inspection system for evaluation of runtime images, for use with a training image that includes a representation of an edge, the training method comprising:
 receiving, with one or more computing devices, an input designating a starting indicator that corresponds to a discrete portion of the edge on the training image;   analyzing, with the one or more computing devices, the training image at the starting indicator to identify one or more characteristics of the edge; and   analyzing, with the one or more computing devices and based on the one or more characteristics of the edge, one or more portions of the edge that are spaced apart from the starting indicator to identify a candidate edge profile for the edge.   
     
     
         22 . The method of  claim 21 , wherein the starting indicator includes an indicator shape that surrounds the discrete portion of the edge on the training image. 
     
     
         23 . The method of  claim 22 , wherein substantially all of the edge on the training image is located outside the indicator shape. 
     
     
         24 . The method of  claim 21 , wherein analyzing the one or more portions of the edge to identify a candidate edge profile includes:
 identifying an expected edge direction based on a first analysis of a first region of the edge;   analyzing a second region of the edge, adjacent to the first region, based on the expected edge direction, to identify a local edge direction at the second region of the edge; and   if the local edge direction at the second region of the edge is not found to correspond to the expected edge direction, analyzing the second region of the edge based on one or more updated expected edge directions that deviate from the expected edge direction by a predetermined angle.   
     
     
         25 . The method of  claim 24 , wherein analyzing the one or more portions of the edge to identify the candidate edge profile includes skipping to a third region adjacent to the second region if attempts to identify the local edge direction at the second region are unsuccessful. 
     
     
         26 . The method of  claim 25 , wherein a number of skips to adjacent portions of the edge based on unsuccessful attempts to identify local edge directions is limited by a predetermined maximum number of skips. 
     
     
         27 . The method of  claim 26 , wherein an end of the candidate edge profile is identified based upon reaching the predetermined maximum number of skips. 
     
     
         28 . The method of  claim 21 , wherein analyzing the one or more portions of the edge to identify the candidate edge profile includes:
 analyzing the edge along a first direction away from the starting indicator until a first end of the candidate edge profile is identified; and   analyzing the edge along a second direction away from the starting indicator until a second end of the candidate edge profile is identified.   
     
     
         29 . The method of  claim 21 , wherein the candidate edge profile is a first candidate edge profile;
 wherein the one or more portions of the edge are analyzed to identify the first candidate edge profile using a first set of analysis parameters; and   wherein the training method further includes analyzing the one or more portions of the edge to identify a second candidate edge profile based upon the one or more characteristics of the edge and a second set of analysis parameters that are different from the first set of analysis parameters.   
     
     
         30 . The method of  claim 29 , further comprising:
 presenting the first and second candidate edge profiles to a user for selection of one of the first and second candidate edge profiles to be provided to an inspection module.   
     
     
         31 . The method of  claim 29 , wherein the first and second sets of analysis parameters include parameters for caliper groups for analysis of the one or more portions of the edge. 
     
     
         32 . The method of  claim 31 , wherein the parameters for the caliper groups include one or more of: a number of calipers per group, a width of individual calipers, a spacing between adjacent calipers, image filter parameters, a contrast threshold, or a caliper extension beyond an expected location of the edge. 
     
     
         33 . The method of  claim 21 , wherein the one or more portions of the edge are analyzed to identify the first candidate edge profile using a first set of analysis parameters, including parameters for caliper groups for analysis of the one or more portions of the edge; and
 wherein the parameters for the caliper groups are determined based on analysis of the edge at the starting indicator.   
     
     
         34 . The method of  claim 21 , further comprising:
 receiving a user input designating one or more regions on the candidate edge profile for masking during subsequent inspection of edges.   
     
     
         35 . The method of  claim 21 , wherein the one or more characteristics of the edge include at least one of: edge color, local edge direction, and minimum edge contrast. 
     
     
         36 . A method for a training an inspection system for evaluation of non-training images, for use with a training image that includes a representation of an edge, the training method comprising:
 identifying, with one or more computing devices, a starting region on the edge on the training image based upon a user-designated starting indicator;   analyzing the edge within the starting region, with the one or more computing devices, to identify one or more characteristics of the edge;   analyzing the edge outside of the starting region, with the one or more computing devices and based on first analysis parameters and the one or more characteristics of the edge, to identify a first candidate edge profile;   analyzing the edge outside of the starting indicator, with the one or more computing devices and based on the one or more characteristics of the edge and second analysis parameters that are different from the first analysis parameters, to identify a second candidate edge profile; and   determining, with the one or more computing devices, an inspection edge profile based on at least one of the first candidate edge profile and the second candidate edge profile.   
     
     
         37 . The method of  claim 36 , wherein analyzing the edge outside of the starting indicator further includes:
 identifying an expected edge direction based on a prior analysis of a first region of the edge;   analyzing a second region of the edge, adjacent to the first region, based on the expected edge direction, to identify a local edge direction at the second region of the edge; and   if the local edge direction at the second region of the edge is not successfully identified based on the expected edge direction, analyzing the second region of the edge based on one or more updated expected edge directions that deviate from the expected edge direction by a predetermined angle.   
     
     
         38 . The method of  claim 37 , wherein analyzing the edge outside of the starting region further includes:
 if the local edge direction at the second region of the edge is not successfully identified based on the one or more updated expected edge directions, analyzing a third region of the edge adjacent to the second region, based on the expected edge direction.   
     
     
         39 . The method of  claim 38 , wherein analyzing the edge outside of the starting region further includes:
 if the local edge direction at the third region of the edge is not successfully identified based on the expected edge direction, analyzing the third region of the edge based on an additional one or more updated expected edge directions that angularly deviate from the expected edge direction by the predetermined angle, as measured from the first region of the edge.   
     
     
         40 . A training system for edge inspection comprising:
 an imaging device configured to capture a training image of an edge; and   a processor device and a memory configured to store executable software, wherein the executable software includes instructions, executable by the processor, for:   receiving the training image;   receiving a user input that provides a starting indicator designating a discrete portion of the edge on the training image, with substantially all of the edge on the training image being located outside of the discrete portion;   analyzing the edge within the discrete portion of the edge to identify one or more characteristics of the edge;   analyzing the edge outside of the discrete portion of the edge, based on the identified one or more characteristics of the edge, to determine one or more candidate edge profiles.