Automated coin grading system

• US 5,224,176 A
• Filed: 07/01/1992
• Issued: 06/29/1993
• Est. Priority Date: 02/22/1991
• Status: Expired due to Fees

• Current Assignee: Professional Coin Grading Service
• Original Assignee: Professional Coin Grading Service
• Inventors: Crain, Louis M.
• Primary Examiner(s): Couso, Jose L.
• Application Number:
  US07/907,239
• Time in Patent Office: 363 Days
• Field of Search: 382/1, 382/7, 382/30, 382/34, 382/37, 382/48, 356/394, 356/445, 356/371, 358/101, 358/106, 358/107, 73/163
• US Class Current: 382/136
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 7/001 using an image reference approach
• CPC Class Codes: G06T 7/33 using feature-based methods
• CPC Class Codes: G06T 7/73 using feature-based methods
• CPC Class Codes: G07D 5/005 Testing the surface pattern, e.g. relief
• CPC Class Codes: G07D 5/10 Testing the rim, e.g. the milling of the rim

Abstract

A method and system for automated grading of coins involves alignment of a target image of the coin with a reference image by mechanically moving the coin until the target and reference images are aligned, capture of multiple images of the coin under a variety of lighting conditions, and subsequent grading using an expert script made up of individual directives which can be arranged to form different scripts for different grading standards and types of coins to be graded. The system and method provide for identification of the criteria used by an expert in grading an object such as a coin, capture of those criteria, and applying the criteria to the grading of the coin in a manner which is implementable on low-cost and widely available microcomputer hardware. The process by which the system is used to identify and test the criteria for evaluating the coin is interactive, and thus, structural changes need not be made in the system in order to grade the different types of coins or apply different coin grading standards.

Claims

1. A method of developing a system for automated grading of a coin image, comprising the steps of:

a. interviewing at least one expert numismatist to determine the process by which the numismatist grades coins;

b. providing directives each of which contains a command to form an electronic imaging processing or scoring operation using appropriate corresponding variables, and storing said directives in a computer memory;

c. retrieving said directives from said memory and arranging a set of selected directives into a script based on interviews of said numismatist;

d. testing the script by causing a computer to execute the directives in the script in order to electronically process an electronic image of a coin and thereby grade the coin;

e. comparing the grade obtained by said computer upon executing the directives in said script with a grade determined manually by said numismatist;

f. varying said script by retrieving said directives from said memory and arranging a second set of selected directives into a second script;

g. repeating steps d and e.

2. A method as claimed in claim 1, wherein step e comprises the step of using a script to process and grade a large number of randomly sampled coins previously and independently graded by a plurality of numismatists to provide subgrades for each of the characteristics known to effect the grade of a coin and step e comprises the step of analyzing results of said comparison in constructing a set of target answers based on a weighted averaging of the results of said previous and independent grading by a plurality of numismatists.

3. A method as claimed in claim 2, wherein step d further comprises the step of using a step-wise linear regression to define optimum weights between the metrics so as to maximize the correlation value of the prediction to the target values across a set of said sampled coins.

4. A method as claimed in claim 1 wherein step e further comprises the steps of:

using multi-dimensional optimization to maximize correlation to the target answers by defining in a spreadsheet the methodologies for combining the results of electronic scoring operations, said results defining metrics;

placing the results into the spreadsheet;

defining initial values for a set of parameters;

using a first parameter to define a table of correlation results as that parameter is varied form one value to another in a series of steps;

assigning the first parameter a value within a range that maximizes correlation; and

repeating the same process for a second parameter.

5. A method as claimed in claim 4, wherein an optimum starting location for multidimensional optimization is defined by using genetic algorithms.