Ping Gallivan

1
Automated Coin Grader

• Ping Gallivan
• Xiang Gao
• Eric Heinen
• Akarsh Sakalaspur

2
Overview

• Introduction
• Technical report
• -histograms
• -edge Detection
• -web Interface
• Conclusion
• Demo

3
Long Term Goal of Project

• Develop a system that will be used to grade,
  appraise and authenticate valuable collectibles
  items such as rare coins providing consistent and
  repeatable results.

4
The Need for anAutomated Coin Grader

• Unreliable results from manual grading
• Value of the coins
• Grading judgment changes from person to person
• Fakes are plentiful
• Many different denominations of coins
• The rare coin market is dynamic and with
  significant changes occurring every week or so

5
Goals for our project

• Develop an Automated Coin Grading System
• Web based Coin Grading Quiz

6
Tools Used

• Java
• Java Script
• HTML
• C
• Imaging Processing Packages

7
Architectural Designof System Overview
Scanner
Image Processor
DB
Extracts features
Scans
Output
System
Display
Grades
8
Creating Database

• Obtain a Coin Image (.gif)
• 36 Coins Histograms
• 36 Coin Edge Detection Images
• Distance Measurements

9
Image Processing

• Hue the color reflected from or transmitted
  through an object.
• Saturation Saturation- the strength or purity of
  a color
• Brightness Brightness- the relative lightness or
  darkness of a color

10
Image Processing Measure Histogram Obtain
statistical data on the scanned pixels in the
image in terms of the Hue, Saturation
Brightness vectors
11
Distance MatrixThe statistical data collected
in step 2 allows us to determine which coins are
similar to others in our database in terms of
known grade.
12
Histrogram Analysis
13
Coin Grade Processing Results
14
Web Based Coin Grading Quiz Site
15
Benefits of the Quiz Site

• Educate and attract new collectors with a fun
  and interactive web interface
• Acclimate the public and the coin grading
  industry to the idea of electronic grading

16
Image Processing Edge Detection


Edge Detection allows us to look at a coin in a
3D view and pickup additional features.
17
Web Site Results Page
18
Analysis

• ..nothing can compare to examining a coin in
  person.
• Four distinct factors
• Surface Preservation
• Strike
• Lustre
• Eye-Appeal

19

• Surface Preservation - This includes the presence
  of bagmarks, hairlines from cleaning or
  mishandling, and other imperfections, whether
  mint caused or man made.
• Strike - Refers to the sharpness and completeness
  of detail, with the normal characteristics of
  that particular type, date and mint mark taken
  into account.
• Lustre - This encompasses the brilliance, sheen
  and contrast of the coin, again taking the normal
  characteristics of the particular issue into
  account
• Eye-Appeal - That certain aesthetic appeal that
  results from the combination of all of the coin's
  qualities.

20

• Process
• Single image of the coin under defined lighting
  conditions should be captured in digital form
  using a high resolution camera.
• Various portions of the captured images are to be
  computer enhanced to bring out important features
  of the coin.
• The key regions of the coin need to be examined
  in great detail to identify, classify, measure,
  and score all flaws.
• A light flow and reflectance analysis should be
  used to precisely measure the mirror as well as
  the inherent lustre of the coin.

21
Future Work

• Expand image processing to include advanced
  feature recognition beyond HSB and Edge
  Detection.
• Increase the database to include a larger sample
  set and other denominations.
• Design an intuitive user interface for scanning
  and grading.
• Move closer towards automated grading
• Secure funding to cover the costs of equipment
  software required

22

• Future Work
• Key components of the coin including obverse and
  reverse marks, strike, lustre, eye appeal,
  mirror, toning, and exceptional conditions need
  to be considered to arrive at a set of expert
  rules.
• Expert Rules Final Grade

23
Conclusion

• what does the future have in store for the
  grading of coins?
• Aid the human graders in making a final
  determination of the grade of the coin
• Computer grading systems can be highly
  consistent, accuracy of about 90
• Image archiving will store one or more images of
  the coin for future reference
• Reduces turn around time and cost

24
Questions???Demonstration