CodeSure TM Printed Code Verification and Inspection ... CodeSure White Paper 4-17-08.pdf ·...
Transcript of CodeSure TM Printed Code Verification and Inspection ... CodeSure White Paper 4-17-08.pdf ·...
Machine Vision Consulting
CodeSureTM Printed Code Verification
and Inspection Technology
Written by:
April 15, 2008
Machine Vision Consulting
7 Old Towne Way Sturbridge, MA 01518
Contact: Joe Gugliotti Cell: 978-551-4160
Fax: 508-347-1355
[email protected] www.machinevc.com
Machine Vision Consulting
Machine Vision Consulting Overview
Machine Vision Consulting, Inc. is based in Sturbridge, MA and has a lab building in
Westborough, MA and an assembly building in Troy, NY. MVC is focused on the
integration of machine vision technology to provide automated inspection and
process control during the manufacturing and packaging processes in a wide variety
of industries.
• End users come to MVC for complete machine vision solutions.
• Machine builders and automation integrators work with MVC to develop
the machine vision portion of their overall assembly, processing, handling, or
packaging solution.
• OEM’s of packaging systems, code printers, robotics, and other process
systems work with MVC as an extension of their engineering organizations to
design, install, and support vision system options.
As the vision industry has matured with easier-to-use products, one thing remains
the same - vision projects are inherently complex. The development and deployment
of a vision system requires a team of experienced vision engineers that can avoid
potential problems that arise when combining high technology from multiple domains
(PLC communications, robotics, vision architecture, real world lighting and optics,
motion, human intervention).
MVC works with its clients to provide a thorough evaluation of the application, a
detailed proposal with images, and a complete vision solution per the specifications
and system acceptance criteria. Machine Vision Consulting works with the client as a
partner and provides sound guidance to assure the application is done correctly the
first time.
MVC has developed turn-key solutions based on its machine vision experience and
expertise. One of these solutions is verifying the correctness and legibility of codes
that are printed onto the ends of cans, the bottoms of aerosol containers and bottles,
packages, labels, or lids. The CodeSureTM solution is discussed in this paper.
Introduction to Code Verification
The focus of this white paper is on using machine vision for the verification and
quality inspection of alphanumeric codes that have been printed onto a can end, the
bottom of an aerosol container, a bottle bottom, a flat container, or a jar lid. By
verification, the focus is that the vision system will automatically confirm that the
expected code has been applied to the container and is legible. The code may
contain a lot code identifier, an expiration date, a plant code, or some other
important production information. Any container that does not contain the expected
code information or that is determined to contain illegible print will be kicked off the
production line.
The printing method often used to create the code is a Continuous Inkjet Printer.
While this method generally produces an acceptable code visually, the printing
process has some inherent variations that make the application of automated code
verification more challenging.
Machine Vision Consulting
Why Is Code Verification Important?
Code verification facilitates the execution of a product recall. If the printed code is
incorrect, missing, or illegible, the producer has no way to verify the pedigree of the
product, nor does the consumer. All traceability to a lot code and expiration date is
lost once that product leaves the printer with an illegible, incorrect, or missing code.
The lack of codes has caused some producers to be forced to recall everything made
because un-coded product has made it to the consumer.
Code verification is critical for bright-stocking. Unlabeled, but coded, generic product
is sent to a warehouse for storage until a customer requests that particular product.
The product is then brought into a labeler and code verification is required to assure
the consumer that the appropriate product container receives the appropriate label.
This is especially important for contract packagers. Cans are blind items and the
printed code provides the only way to know what is supposed to be inside. Vials may
contain a clear liquid that cannot be identified by looking at it. Bottles with tamper
seals all look the same in their unlabeled form except for the printed codes that may
be on the bottom.
Code verification is important to consumer safety. In that cans are blind items,
mixing up a soy product with a milk product or mixing clam chowder with chicken
soup can lead to serious consequences to the consumer and to the packager. If
drugs are involved, correct product identification becomes critical and contributes to
E-Pedigree conformance and patient safety.
In addition to characters that are acceptably distorted, incorrect characters may be
present. If an incorrect code is being printed and the printing error is not detected
quickly, the loss can be substantial, perhaps an hour or two of production that must
be destroyed because its history cannot be verified. Once a product is marked
incorrectly, it generally has to be destroyed, as its pedigree can no longer be
guaranteed.
Given these issues, the machine vision software that is used to verify these codes
becomes the critical component of a code verification solution. If a character is so
distorted that it is unreadable, the code needs to be rejected. However, an
acceptable level of character distortion must be tolerated by an automated
verification system to minimize false rejects and maintain the yield of the line.
How And Why Do Printed Codes Vary In Quality?
It is normal for many on-the-fly code printers to suffer random variations in the
appearance of their printed characters; there are many process setup variables, such
as line speed variations, marking head misalignment, package shape, nozzle
distance, and other factors. Thus, the yield of the vision system used to verify the
accuracy of these critical printed codes is a key contributor to the yield of the line –
how often you get the correct answer, whether it’s Pass or Fail, even when the
individual characters or the entire string varies in appearance because of processes
one cannot always control.
Machine Vision Consulting
Appearance variations are based on the moment-to-moment fluctuations in the
quality of the ink jet-printed codes. This quality fluctuation is often a result of the
human interaction with the print head set-up, especially after cleaning or servicing.
The print head must be installed straight or slanted print may result. The distance
between the print head and the object must be consistent between runs or the size
of the characters can change. If the container is rotating during printing, the printed
code will reflect that in its appearance, creating bowing. In the worst case, the print
head is clogged or is out of ink, creating a situation of “no code” or illegible codes.
Skew Bow Perspective
As character formation variations are brought into the application, the challenge is to
interpret what characters have been printed despite acceptable variations such as
bowing, skewing, stretching, compressing, or flexing. Truly defective codes must be
rejected, while an automated code verification system must accept codes that may
not look perfect, but are acceptable when the code content is correct and legible.
Dealing With Character Distortion And Rotation
An advanced vision software tool from CognexTM, called OCVMaxTM, forms the
foundation for an Optical Character Verification solution from MVC called CodeSureTM.
CodeSureTM imports the appropriate electronic font file to eliminate the need to have
the operator train the system as to what characters look like. The vision system is
verifying that the expected characters are present, so it understands what the
characters should look like in the image based on the importation of the font file
used by the CIJ to create the printed code.
CodeSureTM can tolerate random character quality and appearance variations that
must be tolerated on the production line in order to keep the line efficiency high.
CodeSureTM can also verify rotated codes, critical when round containers are being
verified. In this case, the codes can often be presented in any theta orientation.
Machine Vision Consulting
Font and Character Training
With lesser solutions, the characters of the font being used are trained by taking
images of every character under varying conditions and building up a statistical
model of what each character looks like. In an instance where the characters exhibit
moment-to-moment variations in appearance, this image-based font training would
be a never-ending process.
With CodeSureTM, the same electronic font file that the code printer uses is imported
into the CodeSureTM solution. A TrueTypeTM font is an example of an electronic font
file. Most ink-jet printing systems are utilizing an electronic font description file and
CodeSureTM uses the same file to train the shapes of the character models.
Font training involves selecting a font file to import into the vision system, making
the trained characters in the vision system look the same as what the characters
printed should look like. At run time, the CodeSureTM algorithms compensate
automatically for any acceptable variations in the appearance of the characters in the
printed code.
The font file import function contributes to robustness, ease of use by the operator,
and the maintenance of system validation.
CodeSureTM automatically loads the following font file types:
Xymark – Simplex, Simplex A, and Simplex Roman using .vf or .vb files.
Domino – Arial, OCR-A, OCR-B, and Roman using .cst files.
Videojet – 5X7, 7X9, 10X16 using .xcl files.
Markem – 5X5, 5X7, and 10X16 using .ffm files.
MVC will work with the client to allow CodeSureTM to work with other font files as
needed.
Entering The Code String To Be Verified
For safety purposes, MVC recommends that the character string to be verified is
entered manually into CodeSureTM via the included Operator Interface. While it is
sometimes required that the character string be sent automatically by a line
controller to both the printer and to the code verification system at the same time,
this could potentially lead to the wrong code being printed and the wrong code being
deemed acceptably verified by a machine vision solution.
Some double-check needs to be instituted in this case to prevent an incorrect code
from being printed and accepted. For example, this automatic data entry from the
line controller should at least be done in two separate steps, one for the code printer
and one for the CodeSureTM verifier. The codes should then be checked by a
supervisor to be sure each is correct.
Machine Vision Consulting
Good Character String Candidates
Character strings that can be successfully analyzed with CodeSureTM typically share
most of the following characteristics:
Characteristic Description
Clear Image The image contains sharp character edges.
Good Contrast The image contains a minimum contrast level of 30 grey
levels between the characters and the background.
Print Quality The characters appear with little distortion from their
expected shapes.
Included Font
The characters in the string are composed of a font
installed as part of the CodeSureTM installation and used by
the CIJ printer.
Good Surface
The string appears on a clean flat surface with little or no
qualities that could alter the desired appearance of a
particular character.
The following images contain the characteristics that make them acceptable
candidates for analysis with CodeSureTM:
Flat Package
Machine Vision Consulting
Flat Label Printed With UV Ink and Imaged With UV Light
(Printed by Videojet)
Poor Character String Candidates
Character strings that cannot be successfully analyzed with CodeSureTM, or any
automated verification system, typically suffer from one or more of the following
characteristics:
Characteristic Description
Poor Image The acquired image does not appear in
focus or properly illuminated.
Font Quality
The characters appear distorted, either
throughout the entire string or in relation
to each other.
Font Size
Characters must have a minimum area of
20 x 15 pixels and a maximum area of
100 x 80 pixels. Characters in a string
must be of the same size.
Low Contrast
Images must have a minimum contrast
of 30 grey levels between the characters
and the background.
Poor Surface
Qualities of the surface where the string
appears prevent the string from reliably
appearing with sufficient quality.
Machine Vision Consulting
Insufficient Contrast
The characters in the following image do not show enough contrast:
Label
Unpredictable Surface Area
In the following image, the surface material underneath the character string rotates
randomly during the print application, causing unpredictable character defects:
Plastic Bottle Bottom
Machine Vision Consulting
Poor Font Quality
In the following images, the characters are too distorted to be verified reliably:
Machine Vision Consulting
Overview of the Image Formation Process
An image formation module will be placed either over or under the conveyor
depending on the location of the printer. The imaging module contains the video
camera, optics, and lighting needed to create an image of the code. The containers
will be on the conveyor and will be passed through the imaging module. No container
handling or specific orientation of a round container is required. The containers can
touch during code verification.
An example of a top-side imaging module follows. The type of lighting shown is used
to image jar lids and is designed to minimize the appearance of lid geometry and
color background graphics printed on the lid, while making the code visible. The
camera on the top of the light is verifying the codes on the lid. The lower side-
mounted cameras and reflectors are providing label verification, which is discussed in
another MVC document.
Imaging Module Over the Conveyor, Shown Without Covers.
Example Only. Each Application is Different and May Require a
Different Imaging Module and Lighting Design.
Camera
Light
Conveyor
Machine Vision Consulting
The lighting system is designed to provide a clear image of the jar, can, bottle, or
aerosol container top or bottom and is custom-designed for each application. The
goal of the lighting system is to minimize the appearance of any can ridges,
background graphics, or surface concavity.
A sensor will detect that the container is in position and will trigger the CodeSureTM
system for each container.
The resolution of the camera used will be dependent upon the container speed
required on the conveyor line, the size of the container, the resolution required to
confirm each character, and the variation in code placement on the surface. Cameras
are available from 640 X 480 pixel resolution up to 2448 X 2048 pixel resolution.
A reject mechanism can be mounted on the conveyor after the code is verified, such
that any non-conforming coded containers would be ejected from the line.
Conclusions
The cost of verifying that the correct code has been applied to a container can be
easily outweighed by the cost of a lawsuit or a recall. Safety and allergen prevention
are at the top of the priorities list for many manufacturers. Whether it’s a matter of
life and death or a consumer’s inconvenience, today’s manufacturers are making
sure that their product is accurately coded.
Every application is different and may require a different camera resolution, a
different lighting design, may differ in line speed, may require wash-down capability,
and can vary in many ways.
Additional information on Machine Vision Consulting, Inc. can be found on its web
site, www.machinevc.com. Call Joe Gugliotti at 978-551-4160 or e-mail
[email protected] to initiate a conversation on CodeSureTM.
Some images and descriptions contained herein are from the Cognex Corporation document entitled: “OCVMax Application Guide” and dated June 2007.
About the Author:
Joe Gugliotti is responsible for Sales & Marketing at Machine Vision Consulting and
has been a part of the machine vision industry for 22 years. He spent 10 years with
a leading distributor of cameras, optics, and lighting systems for use in machine
vision applications, working closely with Vision OEM’s, Systems Integrators, and End
Users across the U.S. to evaluate their image formation needs and provide the
appropriate solution. Joe spent the next 10 years as an OEM Account Manager and
Senior Sales Engineer for a leading Machine Vision OEM, gaining exposure to
hundreds of applications across a wide range of industries and using vision
technology ranging from simple sensors to advanced PC-based systems and custom
cutting-edge solutions. In his position at MVC, Joe applies all of this experience to
providing full machine vision solutions and consulting services.