Vision Systems: Seeing the Light
by Alexander Kolchinsky
The perfect vision system
recognizes and rejects every
bad part while accepting
every good part.
Vision systems, sometimes called machine vision, combine cameras, light
sources and a computer into versatile measuring and inspection systems.
These systems are becoming an extremely effective method of assuring process
quality.
Early systems were integrated into packaging lines for optical character
recognition and proved to be a reliable way to check the accuracy of product
codes and label information. Today, high-resolution cameras, advances in
software and imaging processors, and the availability of powerful, inexpensive
compact computers have made vision systems faster and more reliable than
ever. These advances have moved them beyond simple reading tasks. Their
speed, accuracy and ease-of-use have made vision systems attractive for
a variety of manufacturing and industrial inspection applications.
For example, many industries use vision systems to inspect print quality.
These systems scan for color variation, print integrity and registration
accuracy. Print inspection is accomplished by searching for and processing
edge blurs, smudges, scratches, color-density variation and other flaws.
Pharmaceutical manufacturers use automated vision systems to inspect the
color, shape and size of tablets, as well as to detect cracked or broken
tablets and search for stray tablets outside the package.
Advanced vision systems still "read" characters, too. Optical
character recognition systems compare date and lot code on packaging with
reference information entered into the inspection system. Newer systems
can read multiple fonts and recognize alphanumeric codes and broken characters.
Many other types of vision systems are used in electronic component placement,
solder joint inspection, assembly step verification, in-part integrity checking
operations and surface finish inspection. They are also used in metrology
applications to measure the size, location and form of workpiece features.
Consistency is the advantage
Whether they are integrated into the manufacturing operation and used as
in-process inspection systems, or used as post-process inspection systems,
vision systems aim to identify process problems quickly so that corrections
can be made. The systems' primary advantage is their consistency in performing
the inspection task, whether that task be detecting flaws, counting/sorting,
assembly verification or code/character reading. What sets vision systems
apart is consistent part-to-part inspection (free from operator influence)
and the power that consistency gives to manufacturers to ensure that process
aberrations are kept to a minimum.
Along with performance consistency comes reliability. Well-designed vision
systems work 24 hours a day, seven days a week, with very little need for
service downtime.
Some vision systems also offer users another advantage-the ability to automatically
calibrate and check themselves prior to an inspection cycle. Thus, if the
fiber-optic cable controlling the light malfunctions, the image processor
isn't working properly or the light source isn't operating at optimum illumination,
the time to find out is before the production run/inspection cycle begins.
The best vision systems today perform this self-check quickly and inform
the operator that the system is operational before the run begins. If there
is a problem, the diagnostic routine informs the operator through software
prompts and often offers corrective solutions.
Increasing yield through
electronic inspection
When does a company need a vision system? In some cases, customers and/or
regulatory bodies require 100-percent product inspection with documented
inspection results. In other cases, if production quantities are high enough,
a vision system may be justified.
Quality is another determining factor. Quality production is marketable
through products that perform better, last longer and are more attractive
than their competitors. Vision systems provide a means of increasing yield-that
is, the ratio of good parts to bad parts. When a serial defect is spotted,
the system not only recognizes it but stops the work-handling conveyor and
informs the operator of the defect's location and its magnitude. Using the
information, the operator can correct or adjust the manufacturing process
to eliminate the defect.
The yield factor is particularly important in manufacturing industries that
produce large volumes, as in the compact disc and pharmaceutical industries.
By reducing the number of defective products, the system can pay for itself
very quickly, sometimes in a matter of months.
Selecting the appropriate system
If a vision system is right for your organization, what should you look
for when scouting the market? The first step is to consult a recognized,
experienced vision systems vendor. No two vision systems are alike in operation.
Most are modular in design and can be assembled to application-specific
requirements. First, an experienced vendor can help you determine if a vision
system is appropriate for your inspection task and then help you select
the system that best solves your problem.
The most critical consideration, of course, is the system's ability to accurately
and consistently detect product flaws, ignoring normal product variation.
The perfect system recognizes and rejects every bad part while accepting
every good part. Realistically, the vision system still must reject every
bad part while keeping the yield on good parts very close to 100 percent.
State-of-the-art vision systems must present a sufficient history and classification
of defect information to the operator to help identify the process problem
and fix it in as short a time as possible. These systems must also automatically
generate reports and job histories to help the quality assurance supervisor
with off-line analysis.
Systems should allow quick job setup through an interactive operator interface.
The operator should be able to set up the system and fine-tune it for a
specific inspection routine in 60 seconds or less.
The cost of the system should be evaluated in combination with its performance.
In high-volume applications, a more expensive system can easily be justified
by the higher yield of good parts passed.
Contrast is critical
The operating principal of vision systems is to separate the inspected feature
from the background by as many gray levels as possible. The system's sensitivity
depends on the choice of camera and light source, among other things. Both
black-and-white and color cameras can be installed in vision systems. The
choice depends upon the inspection application. If the part being inspected
has colors that are similar in intensity, or the feature to be inspected
has a similar intensity to the part's background intensity, a black-and-white
camera may not provide adequate contrast to spot defects. A color camera
would be a better choice.
LED and halogen light sources can be installed in vision systems. Each has
advantages. An LED light source lasts a long time, exceeding 40,000 hours
in some cases. Its narrow spectrum makes LED light a good choice for black-and-white
camera applications, and, in most cases, it works well with ambient light.
Halogen light is a preferred choice in color-sensitive applications. A halogen
light must be changed more frequently than an LED light but is easier to
change and costs less.
Systems can also be equipped with multiple cameras and light sources, depending
on the spatial resolution-the accuracy-that the system is required to achieve.
However, multiple cameras may also be required if more than one type of
defect is to be found on the same part. Multiple cameras require more processing
time, which could come at a higher cost or slower overall system speed.
Again, the vision systems vendor is the best source of advice on correct
system configuration for a specific inspection application.
Operator interface
enhances system potential
Due to the process-control aspects of vision systems, the operator interface
of the system is critical to performance. Systems with a graphic user interface
are the easiest to use. Operators simply select icons corresponding to the
actions they want to perform. Screen prompts guide them through setup and
analysis. Some characteristics of a good interface are the clarity and simplicity
of defect information presentation and ease of access to on-line program
help. Some systems provide touch-screen graphic user interfaces. The touch-screen
interface is the most reliable system interface because it has no mechanical
parts to break down.
Hardware is another consideration. Select a standard, industrial-grade open-platform
computer, not a proprietary solution.
Another selection factor is how inspection results will be handled. The
system should be capable of communicating with a PLC through parallel I/O
or serial ports.
Vision-system technology is changing rapidly as this approach to inspection
becomes more widely accepted in industrial applications. If a vision system
is right for you, your manufacturing operation stands to gain a significant
improvement in quality production.
About the author
Alexander Kolchinsky has more than 20 years of experience designing and
installing vision and medical imaging systems. He is president of Automated
Visual Inspection Systems, LLC, a joint venture between Autoroll Machine
Co., LLC, and a group of senior image-processing scientists and engineers.
