Color tolerancing system employing fuzzy logic

• US 5,410,637 A
• Filed: 06/18/1992
• Issued: 04/25/1995
• Est. Priority Date: 06/18/1992
• Status: Expired due to Fees

• Current Assignee: Datacolor Holding AG
• Original Assignee: Color and Appearance Technology, Inc.
• Inventors: Kern, Kevin B., Alston, David L.
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US07/900,564
• Time in Patent Office: 1,041 Days
• Field of Search: 395/900, 395/61, 209/580, 364/526
• US Class Current: 706/52
• CPC Class Codes: G01J 2003/466 Coded colour; Recognition of predetermined colour; Determining proximity to predetermined colour
• CPC Class Codes: G01J 3/46 Measurement of colour; Colour measuring devices, e.g. colorimeters measuring colour temperature G01J5/60
• CPC Class Codes: G01J 3/462 Computing operations in or between colour spaces; Colour management systems
• CPC Class Codes: G01J 3/463 Colour matching
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

A color measuring system uses fuzzy logic to establish acceptable pass/fail tolerances for production or inventory samples. The tolerances define a super ellipsoid which has six individual tolerance coordinates. Initially, the tolerances are defined by the high/low values of the pass samples in a training set which have been chosen to represent a population of visually rated pass and fail samples. A classifier template uses the super ellipsoid tolerances and ranks every sample in the training set. Any pass/fail ranking that differs from the visual rank is considered an error. All of the error distances are summed and all samples are passed through the fuzzy logic contained in an induction module to determine the changes needed in the super ellipsoid tolerances to minimize the error distance. The new tolerances are then passed to the classifier template for a new ranking of the training samples. The process is repeated until a minimum error distance is achieved. The final tolerances are then used by the classifier to rank subsequent samples as either pass or fail. Shade sorting can also be accomplished. In that case, the training samples are rated twice, first for the centroid color differences and then for the shade differences. The training set is processed by the classifier template/induction module to determine the two tolerances. The classifier then rates production/inventory samples to determine if they meet the centroid color needs, and then places the samples in the proper shade sort group.

Claims

1. A color measurement system comprising:

comparing means for comparing a plurality of pass/fail samples in a training set against a classifier template based upon a specific geometric volume, said geometric volume comprising a super ellipsoid having at least six different tolerance coordinates; and

,

fuzzy logic means for determining, based upon the sum of error distances of said samples from a reference point, which tolerances to change to minimize unclassifiable tolerances, and for changing at least one of said six different tolerance coordinates of said super ellipsoid in response thereto,

wherein said fuzzy logic means establishes a new geometric volume and a trained classifier template and wherein said trained classifier template is employable to determine the pass and fail classifications of subsequent sample sets.

2. The system of claim 1 wherein said super ellipsoid comprises a geometric volume defined by the following relationships:

##EQU3## where L, C, H represent an input sample, and,

KL₀ is the negative tolerance for L, and,

KL₁ the positive tolerance

KC₀ is the negative tolerance for C, and,

KC₁ the positive tolerance

KH₀ is the negative tolerance for H, and,

KH₁ the positive tolerance and

if L≦

0, KL=KL₀ otherwise KL=KL₁ ; and

,

if C≦

0, KC=KC₀ otherwise KC=KC₁ ; and

,

if H≦

0, KH=KH₀ otherwise KH=KH₁.

3. The system of claim 2 further comprising:

reclassifying means for production samples if said samples fall into a region of uncertainty and for introducing reclassified samples into future training sets to establish new tolerance values for new super ellipsoids and new classifier templates.

4. A color sorting system for sorting colors based upon geometric volumes having centroid colors, said system comprising:

comparing means for comparing a set of samples wherein each sample has two values including a first value which represents the pass and fail values for centroid color and a second value which represents the pass and fail values for shade color, wherein said samples are compared against two classifier templates based upon two sets of tolerances established by two geometric volumes respectively; and

,

fuzzy logic means for changing the tolerances of said two geometric volumes respectively and for determining, based upon the sum of the error distances from a reference point, which of the tolerances to change thereby defining two new geometric volumes based upon the changed tolerances and establishing two new trained classifier templates based upon the two new geometric volumes,

wherein the shade center of the geometric volume covers the entire centroid color geometric volume and wherein the two trained classifier templates are employed to determine the pass and fail classification of subsequent samples and then to determine into which shade groups said samples should be placed.

5. The system of claim 4 wherein said two geometric volumes comprise two super ellipsoids having at least six different tolerances for each coordinate and wherein said two super ellipsoids are each defined by the following relationships:

##EQU4## where L, C, H represent an input sample, and,

KL₀ is the negative tolerance for L, and,

KL₁ the positive tolerance

KC₀ is the negative tolerance for C, and,

KC₁ the positive tolerance

KH₀ is the negative tolerance for H, and,

KH₁ the positive tolerance and

if L≦

0, KL=KL₀ otherwise KL=KL₁ ; and

,

if C≦

0, KC=KC₀ otherwise KC=KC₁ ; and

,

if H≦

0, KH=KH₀ otherwise KH≦

KH₁.

6. An appearance measurement system comprising:

comparing means for comparing a plurality of pass/fail samples in a training set against a classifier template based upon a specific geometric volume, said geometric volume comprising a super ellipsoid having at least six different tolerance coordinates; and

,

fuzzy logic means for determining, based upon the sum of error distances of said samples from a reference point, which tolerances to change to minimize unclassifiable tolerances, and for changing at least one of said six different tolerance coordinates of said super ellipsoid in response thereto,

wherein said fuzzy logic means establishes a new geometric volume and a trained classifier template and wherein said trained classifier template is employable to determine the pass and fail classifications of subsequent sample sets.

7. An appearance sorting system for sorting colors based upon geometric volumes having centroid appearances, said system comprising:

comparing means for comparing a set of samples wherein each sample has two values including a first value which represents the pass and fail values for centroid appearance and a second value which represents the pass and fail values for shade appearance, wherein said samples are compared against two classifier templates based upon two sets of tolerances established by two geometric volumes respectively; and

,

fuzzy logic means for changing the tolerances of said two geometric volumes respectively and for determining, based upon the sum of the error distances from a reference point, which of the tolerances to change thereby defining two new geometric volumes based upon the changed tolerances and establishing two new trained classifier templates based upon the two new geometric volumes,

wherein the shade center of the geometric volume covers the entire centroid appearance geometric volume and wherein the two trained classifier templates are employed to determine the pass and fail classification of subsequent samples and then to determine into which shade groups said samples should be placed.