Electro-optical quality control inspection of elements on a product

Claim 2

Apparatus for optically inspecting a graphic pattern on a product for identity with the graphic pattern on a previously-examined training sample, comprising: a video camera, means for positioning a product to be inspected so that said camera scans a rectangular frame portion containing graphic patterns of said product, window gate means operative to define a plurality of electronic window areas within the scanned frame, and to pass the video signal from the camera solely during scans through said window areas, means to manually preset the sizes and locations of said windows so that they closely surround the training graphic pattern, means operative during an inspection of a product to automatically enlarge said windows enough to be sure that said graphic patterns are included therein, and then contract said windows until they closely surround said graphic pattern, said means to enlarge each of said windows including means to compute a plurality of vertically-enlarged overlapping search windows and a plurality of horizontally-enlarged overlapping search windows, video signal processing means operative to translate the video signal passed by said window gate means into representative feature sets of digital signals, and means to compare said feature sets of digital signals with feature sets of digital signals obtained during a previous examination of the corresponding rectangular frame portion of the training sample.

Claim 3

The combination as defined in claim 1 wherein said means to contract said windows includes means to find the vertical position of the graphic pattern in the vertically enlarged search windows, and means to find the horizontal position of the graphic in the horizontally-enlarged search windows.

Claim 4

The combination as defined in claim 2 wherein said means to contract said windows includes means to generate a run-length-code signal representing the horizontal scan lines in which a graphic pattern is encountered.

Claim 5

The combination as defined in claim 3 wherein said means to find the vertical position of the graphic pattern includes means to figuratively move a row run-length-code signal representing a graphic pattern obtained from the prototype vertically to a position of match with a row run-length-code signal obtained from the graphic pattern in the vertically-enlarged search windows.

Claim 6

The combination as defined in claim 4 wherein said means to contract said windows includes means to generate a column run-length-code signal representing the vertical columns in which a graphic pattern is encountered.

Claim 7

The combination as defined in claim 5 wherein said means to find the horizontal position of the graphic pattern includes means to figuratively move a column run-length-code signal representing a graphic pattern obtained from the training sample horizontally to a position of match with a column run-length-code signal obtained from the graphic pattern in the horizontally-enlarged search windows.

Claim 8

The combination as defined in claim 6, and means to compute a match quality number for the degree of match of the product pattern at the most likely location, and to compare the match quality numbers with a threshold to determine whether the product pattern is similar to the prototype pattern.

Claim 9

The combination as defined in claim 6, and means utilizing said positions of match to create an electronic window closely surrounding the graphic pattern on the product for use in obtaining said graphic feature signals for comparison with graphic feature signals obtained from the training sample.

Claim 10

The combination as defined in claim 8, and means to compute the size of the graphic pattern on the product, and means to compare the computed size with the size of the graphic pattern on the training sample to see if they match.

Claim 11

Apparatus for optically inspecting a graphic pattern on a product to see if it matches the graphic pattern of a previously examined training sample, comprising a video camera, a means to position a training product and later a production product, so that the camera scans a rectangular frame portion on the product, window gate means operative to define an electronic window in the frame, and to pass the video signal from said camera solely during scans through said window, video signal processing means operative to translate the video signal passed by said window gate means into row signals including in a first category signals representing the number of each row scan line of vertical edges V, and in a second category signals representing the number of horizontal down edge points Hdn, and the number of horizontal up edge points Hup, and into column signals including in a third category signals representing the number in each column of horizontal edge points H, and signals in a fourth category representing the number of vertical down edge points Vdn, and the number of vertical up edge points Vup, means to compare the V signals, and Hdn and Hup signals, the H signals and the Vdn and Vup signals, obtained from scanning the product pattern, with corresponding signals obtained from scanning the training pattern, said signals in each of the four categories of signals being the averages of signals obtained over a plurality of frame scans, and means to data-reduce said signals by averaging the signals in the first category V, and averaging the signals in the second category Hdn and Hup, over a plurality of successive rows, and by averaging the signals in the third category H, and averaging the signals in the fourth category Vdn and Vup, over a plurality of successive columns.

Claim 12

Apparatus for optically inspecting a graphic pattern on a product to see if it matches the graphic pattern of a previously examined training sample, comprising a video camera, a means to position a training product and later a production product, so that the camera scans a rectangular frame portion on the product, window gate means operative to define an electronic window in the frame, and to pass the video signal from said camera solely during scans through said window, video signal processing means operative to translate the video signal passed by said window gate means into row signals including in a first category signals representing the number of each row scan line of vertical edges V, and in a second category signals representing the number of horizontal down edge points Hdn, and the number of horizontal up edge points Hup, and into column signals including in a third category signals representing the number in each column of horizontal edge points H, and signals in a fourth category representing the number of vertical down edge points Vdn, and the number of vertical up edge points Vup, means to compare the V signals, and Hdn and Hup signals, the H signals and the Vdn and Vup signals, obtained from scanning the product pattern, with corresponding signals obtained from scanning the training pattern, said signals in each of the four categories of signals being the averages of signals obtained over a plurality of frame scans, and means to data-reduce said signals by averaging the signals in the first category V, and averaging the signals in the second category Hdn and Hup, over a plurality of successive rows, and by averaging the signals in the third category H, and averaging the signals in the fourth category Vdn and Vup, over a plurality of successive columns.

Claim 13

The combination as defined in claim 10 wherein said video processing means includes a spatial filter receptive to the video signal, the spatial filter including, in sequence, a non-linear, five-point median filter to eleminate spike and speckle noise, and a linear filter performing a three-point weighted sum of 3 adjacent pixels to aid in smoothing the signal for reliable edge detection.

Claim 14

The combination as defined in claim 10 and means to compute a transform of the signals in each of the four categories.

Claim 15

The combination as defined in claim 12 in which said transform is a Walsh transform.

Claim 16

The combination as defined in claim 12 and means to compare the low-sequency components of each of the four transforms with corresponding components from the training sample to determine whether the product graphic pattern matches the training graphic pattern.

Claim 17

The combination as defined in claim 14 and means to compute the peak light signal P L and the peak dark P D signal from the video signal, and means to compare said signals with corresponding signals obtained from the prototype.

Claim 18

The combination as defined in claim 12 and means to scan the same pattern on a plurality of training sample products, and to compute a prototype which is an average of the feature signals obtained from all the training samples, for comparison with the feature signal obtained from the production product.

Claim 19

The combination as defined in claim 16, and means to compute a match threshold which is more distant in feature space from the prototype than a multiple of the distance of the most distant training sample from the prototype.

Claim 20

The combination as defined in claim 17, and means to preset the above multiplying factor for each pattern, to provide a gradation in steps from relatively loose to tight threshold tolerance.