Machine Vision Technology Take-Up in Industrial Applications

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Machine vision technology take-up in industrial applications

Anni soini
Satakunta polytechnic
antti.soini@,cc.spt.fi
Abstract

2. Machine vision market and trends

Machine vision technology entered industrial


application already in early 70's. but due to high
computing costs penetrations has been limited.
Machine vision business is pretty fragmented and the
small technology companies are seeking repeat business
in narrow business segments. Many new application
areas with risks involved remain unsolved, because
technology companies are not willing to carry the risks
alone and the endusers are used to buy proven
technology to their production lines.
European Commission funded Information Society
Technologies (ISr) programme has been active in
supporting machine vision technology transfer, take-up
and research. This promotion has encouraged end-users
and technology companies to work together in pilot
application and share the risks. Parallel European-wide
dissemination efforts have carried these solutions to wide
audience ofpotential new users.

Machine vision business continues its strong annual


growth [I]. Based on the research of AIA (American
Imaging Association) [2] in 1999 global machine vision
business reached USD 5 billion milestone. North America
covered 34%, Japan 3 1%, Europe 24% and the rest of the
world 11%. The growth fom previous year was almost
20% and for instance in Germany they estimate over 20%
growth reaching to 450 Million in 2000. In Finland we
have over 50 machine vision companies and they
produced almost 90 million sales last year.
What have been remarkable in this business are joint
ventures and alliances with which many companies have
opened new markets or strengthened their resources and
product portfolios. Still the market is pretty fragmented
and built of too small enterprises, who build their products
based on their own innovations rather than utilising those
widely available in the market.

2.1. Trends
1. Introduction
Faster.. .cheaper.. .more accurately.. .more reliably are
the requirements set to modern machine vision solutions.
These requirements go against each other but the current
development is bringing those features closer to each
other.
Dedicated controls are replaced with open
architectures, Windows-based user interfaces start to be a
common practise. Integration of image, movement and
calibration of subsystems to a common functionality
enable effective and accurate use of machine vision in
robotic controls.
The development in hardware and software has been
active in last few years. Many of the camera manufacturers
have introduced their intelligent cameras, where
controlling computer is integrated in the camera body. The
machine vision system now has become an intelligent
sensor, which could be connected with a process control
system via a fieldbus. This way they will take a big share
of conventional sensors in process automation.

Computer vision technology was turned to industrial


machine vision solutions already from early 70's. Many
pioneers have tried to solve industrial challenges with the
existing technology, but due to limited computing power
and high computing costs only few have managed to make
this a business. Even in these days many industries still
hesitate to invest in machine vision technology. On the
other hand technology companies have tried to focus to
narrow niche applications where the repeat business is
possible, but requires global market penetration.
European Commission funded Information Society
Technologies (IST) programme has been active in
supporting machine vision technology transfer, take-up
and research. This promotion has encouraged end-users
and technology companies to work together in pilot
applications. Parallel European-wide dissemination efforts
have carried these solutions to wide audience of potential
new users.

332

CMOS technology is gaining markets from CCD sensor


elements. This trend supports further integration and
robustness of the new machine vision systems.
Order based manufacturing requests high flexibility.
Intelligent and adaptive machine vision systems
(Cognitive vision) will replace most part of the application
specific programming.
Machine vision manufacturers will benefit from the
trend of directing this business to consumer markets. Very
low-cost systems will be available to consumer
applications. Big players like Microsoft and Intel are
coming to image processing markets with their public
domain software products directed to consumer markets.
Low-cost intelligent cameras collect information, analyse it
and transfer the results via wireless or solid lines for
further processing.
Cognitive vision is approaching industrial applications,
but a lot of research is till to be made before these
solutions are feasible and commonly in use. The
processing power of the systems has doubled in 18-month
periods and this trend seems to continue.

involved in 173 projects, managed by 21 TTNs all over


Europe. The initiative had an overall budget of 70 million
Euro partly funded by the European Commission and
partly by industry.
Each TTN helped local industry adopt machine vision
technology by helping them set up small projects to
specifically address the introduction of machine vision in
their own enterprises. By helping to finance these projects
the EC helped industry reduce the risk of trying it out, and
kept the costs low.
Helping local industries to understand benefits of
machine vision was only a starting point of the TTN
network. The key to the TTN network was supplying
European dimension to these projects. TTNs worked
together to spread the results to European-wide audience.
The developed local solutions were "marketed" to other
countries, so that their industries could see how the
technology could be applied to similar sectors.
The TTN network was set up to help European
industry adopt new technology that had the potential to
help industry to improve its competitiveness and
performance. The results have been spectagular and the
impact huge; 59 new products, 47 product enhancements,
22 new services, 20 enhanced services and 15 new
companies. An estimated 1000 man-years of direct new
employment is foreseen between January 2000 and
December 2002. Over 100 examples of the TTN network
resulting in the transfer of technology between member
states.

3. Research and take-up


In European research environment Commission funded
research programmes have brought researchers from
different countries to work together. Information Society
Technologies ( E T ) programme have promoted machine
vision research and in the current 2001 workprogramme
Cognitive vision has got a special attention.
In addition to funding research, EC has also been
concerned of technology transfer from research to
industry. In the HPCN (High Performance Computing and
Networking) Technology Transfer Node (TTN) Network
running between March 1997 and March 2000 the
Machine Vision Sector Group was one of the most active
groups. The HPCN TTN [3] programme has stimulated the
uptake of high-performance computing by industry. This
initiative has specifically targeted smaller organisations,
which can now afford low-cost powerful systems made
out of commodity components such as standard PCs.
Substantial emphasis has been placed on the novel
application of HPCN to problems 'where it was previously
too expensive or immature.

3.2. Integrated machine vision cluster


Integrated Machine Vision Cluster (EUTIST-IMV) [4] is
a new initiative funded by IST programme Take-up actions
and industry. It was opened in October 2000 and it will be
promoted for 30 months. Principal contractors are located
in Finland, Great Britain, Italy and Germany.
Currently 16 machine vision applications are running and
about ten more will be accepted in near future. Many
manufacturing and production processes can benefit from
Integrated Machine Vision. The EUTIST-IMV initiative is
bringing together business and technology to generate
new solutions, which will bring measurable benefits to
European industry.
EUTIST-IMV has partners from a range of industries sawmills and wood processing, glass manufacturers,
textile weavers, papermills, electronics, food and sheet
metal industries and vehicle manufacturers among others.

3.1. Network of technology transfer nodes


The HPCN TTN initiative was targeted at Small and
Medium sized Enterprises. 609 companies have been

333

. monitor sawblade condition and to choose blade


materials and thickness;
. monitor condition of reducers;
. monitor timing and play of upper and side rolls;
. monitor play of sawing machines and observe effects of
alignment and maintenance activities.
By using the Seecon measurement and reporting
system, a sawmill can tolerate an increase of 10% in raw
material prices or a drop of 5% in sales prices and still get
its previous profit. With this new saw controls a sawmill
producing 120,000 4 of finished timber can increase its
profitability by 1 million Euros per year, assuming no
changes in material costs or sales prices.

They all have one thing m common: they aim to become


more competitive by using innovative new technology.
EUTIST-IMV is open to all participants from industry.
It runs an information service to allow industrial partners
to find out how they can benefit from machine vision, and
a programme of workshops, seminars and other events.

4. Some applications
The programmes above have disseminated results of
many different machine vision applications. Also different
technologies have been applied starting from B/W matrix
camera solutions to high definition colour line scan camera
applications and further more to spectral imaging and
thermal imaging applications [ 5 ] .
In this paper a closer look is made to wood and glass
processing applications. The problem setting, solutions
and industrial benefits are introduced.

4.1. Wood processing


Wood is a natural material with many variations in
shape, structure and quality. Optimal use of this relatively
expensive material offers many challenges to machine
vision technology providers. Applications start from
forest inventory to cutting optimisation, log sorting,
optimal positioning for breakdown saws, dimension
control in cutting processes, quality grading of green
wood and again dry wood. After sawmills the wood
elements (boards, planks, etc.) Continue to further
processing, where again optimal use of wood is the key.
Following example is from breakdown sawing.
4.1.1 Advanced vision systems are used to control the
accuracy of production in sawmills [7].
Control of sawblade behaviour during sawing
determines the yield and profitability of sawmills. Until
now, saw operators have had no real-time means of
getting information about the cutting process. Only
occasional manual spot checks were possible. A unique
measurement system with special optics and software
called Seecon has been developed. It enables real-time
control by measuring and analysing every piece of timber
sawn immediately after cutting. Seecon enables sawmill
operators to:
. adjust sawblade setups;
. optimise line speed and sawblade changes based on wear
and durability;

Fig. 1. Seecon breakdown cutting control.


4.1.2. Technical Overview

At any sawmill, the traditional method of setting up


saws is through manual measurement. When the
measurements seem right, production sawing is started.
However, very few sawmills have the time or interest to
measure the timber produced adequately. They do not
even have the time to analyse even a few hand-made

334

measurements, other than in assessing the order of


magnitude of the cut dimensions. Consequently a new
system for automatic control and measurement has been
developed.
The system developed is a high speed, accurate realtime vision system with flexible and modem
communication capabilities. It enables sawmills to
supervise and control the production and output of
breakdown saws. A breakdown saw is a saw that cuts a
log into timber. A log can typically be cut into four boards,
but it can be cut into as many as 15 boards. The control
system comprises an on-line machine vision unit at the
saw machine in which the data is digitised and sent to a
powerful computer to process the data in real-time. The
speed of a log at the cutting process is typically 1.3 metres
per second. This means that a new log has to be
measured, analysed and reported approximately every
three seconds. The computer is linked to the saw machine,
the sawmill computer network and saw operator for
immediate reporting and statistical control. With this
system, a sawmill can reduce waste of raw material,
increase yield and improve product quality.

A large number of end-user requirements were


determined, because production practices differ from
country to country. Some of the sawmills saw customised
dimensions in small lots and, at some sawmills, the
incoming logs could be of any size. Other requirements
included the need for an automated board size
determination function, communication with other process
computers, long period statistical follow-up, printed and
stored reports, on-line after sales service and flexibility of
installation in vertical and horizontal positions.
4.1.3. The benefits

Previously, after a set-up change was made, a few logs


were sawn, then the line was stopped and measurements
of dimensions were performed manually. After necessary
adjustments, a few more logs were sawn and the
measurement was repeated. This cycle was repeated until
the set-up was right and production couId be started.
When Seecon was brought into use, the set-up was done
on-line and in real time with a consequent significant
increase in production time. The time saved enables the
sawmill to be more productive and so more profitable.
Even if a cost-benefit study is based on average values,
the faster set-up times indicate significant eamings to the
mill of over 600,000 Euros per year.
Another interesting calculation is the cost of inaccurate
(and therefore rejected) boards sawn during set-up when
manual measurement was used. On average it took 20
minutes to make all adjustments for a new set-up. During
this time the mill sawed an average of 30 logs yielding an
average of 90 boards. The sawmill is operated in two shifts
and on each shift, the set-up is changed four times. The
quantity of uncontrolled boards per day was therefore 720
when using the manual set-up procedure. Over a year this
would represent 165,000 uncontrolled hoards half of which
with a value of more than 300,000 h r o s would not be
sufficient quality. If these inaccurately sawn boards were
sold to a customer who complained, the sawmill would
have to reduce the price of whole order by around 30%.
This increases the monetary loss and the suppliercustomer relationship suffers.
Seecon can also improve a sawmills yield and cost
effectiveness in other ways. Production planning can get
feedback concerning the correctness of cutting patterns
and log sorting and the sawmill can use tighter tolerances
that increase yield or chip quantity. Furthermore it is
possible to monitor individual sawblades and to decide
sawblade changes based on wear and tear and not by the
clock. For the first time maintenance personnel can get

Approach
A preliminary prototype to measure board dimensions
in a sawmill had been developed by Ekspansb
Engineering and installed in Finland. This prototype was a
stand-alone unit with no communication capabilities. It
had only limited use due to its slowness, inaccuracy, poor
installability and lack of reporting and control capabilities.
Therefore it could not easily be modified or used by other
sawmills. In this development, a major modernisation and
redesign of the prototype was made. This was done in
collaboration with Vanem Timber and involved the
development and testing of a more efficient processor
architecture necessary for the new higher production and
processing speeds and quality control applications in the
sawmills.
The objectives of the development were:
. The installation of a prototype system in a sawmill in
Sweden in order to modify and test the usability and
performance of the system. This was complemented by
study visits to other sawmills to make sure that the
solution developed would be generally applicable for
sawmills at large;
. The measurement of the breakdown process parameters
before the prototype installation, immediately after
installation, and after the learning period;
. The assessment of the short and long term industrial
benefits of the technology developed.

335

real-time information about the behaviour of machinery


and the saw operator can monitor in real-time the
dimensions of timber produced.
For the first time in sawmill history the operator can
start using real-time and statistical control tools for the
cutting processes. The benefits of Seecon for sawmills will
be significant because:
. Saw blade changes can be performed faster than before
because with the continuous saw-cut control final
adjustments can be performed during line start-up and
operation. This up-time increase can be up to 20 minutes;
. A controlled uniform and consistent production allows
usage of smaller tolerances. This improves the yield and
the sawmill does not need to give away any unnecessary
extra material;
. A report of production tolerance errors is immediately
available;
. An alarm can be triggered by an erroneous set-up of the
saw blades, a wedged piece of wood between the blades
or of blade wear and tear.
The elimination of production with wrong dimensions
means important savings and improves productivity. The
software makes possible accurate short and long term
analysis. This is important not only for the production
engineer but also in case of rejected boards. With the
system the sawmill can check each individual boards
production history.
One of the primary reasons for a sawmill to invest in
and to use this new quality control system is because they
can reduce waste of raw material and improve their yield
and quality. The system assists service personnel with the
set-up and control of the saw blades and the cutting
dimension measurement immediately after sawing. A
controlled uniform and consistent production allows
usage of smaller tolerances. The end result is that wood is
used in a more sustainable manner that saves natural
resources and earns more money to the sawmill.

i
Figure 2. Top end wane and log sweep wane.
The novelty is the three-sided measurement in an
edger-saw, and the combined usage of knot and wane
data. The design goal is that the system can be integrated
to existing and new edger-saws. The exploitation
objectives are to replace the existing insufficient
measuring units capable of measuring only the wane with
a more powerful device, and to offer an affordable system
alternative to sawmills and saw machine manufacturers.

4.2. Glass measurement


4.2.1. Quality control of glass using laser interferometry
181
Introduction
PanOptix Oy has developed an innovative, laser-based
system for the inspection and quality control of float
glass. This system can inspect thickness across a sheet of
glass on-line during production and its edge distortion
off-line. The system is controlled by a Windows-based PC
and output is presented graphically. Variations of only 0.3
microns in 10 mm thicknesses can be detected off-line and
of 10 microns across a sheet of width 3.8 m on-line. The
system can automatically detect the borderline between
high quality glass and glass of variable thickness. This
enables optimised cutting with resultant material and
energy savings.
The major benefits of the PanOptix system are:
fully automatic measurement of float glass
quality;
better quality control;
optimised cutting resulting in less waste;
accurate glass thickness measurement;
greater uniformity of glass plates;
elimination of human error in quality control.
The system produces an impressive return-oninvestment and results in an increase in productivity from
2 to 4% on float-glass production lines.

4.1.4. Further development


The Seecon system is further developed to be a realtime colour vision and control system to masure the
features of a board from three sides when it passes
longitudinally through the measuring unit immediately
before it enters an edger-saw. The data on recorded
features is used to analyse what is the best sawing pattern
for the rough board in order to get the best possible value
recovery. After the value optimisation the system sends
the required control signals to the saw machinery.

336

Testimonial
Plant Manager, Reijo Sihto, Pilkington Glass Lahti Ltd,
testifies: Optical off-line measurement has improved our
quality control and process development. The new online device will bring these tools to bear in real-time
process management. We expect to get between 2 to 4%
more high-quality glass from the same line.
4.2.2. Technical Details

On-line measurement of glass thickness giving better


quality with high line yield
Real-time information from the glass production line
using the PANOPTIX ON LINE MM 380lenables the
optimisation of:
glass thickness;
ribbon width;
ribbon profile;
edge distortion.
The main features of this product are summarised in the
table 1. below:
Principle: optical measurement based on laser beam
reflection from the front and back surfaces of
the glass. The accuracy and reliability of the
measurement is enhanced by a specially
designed diffractive element to modify the
laser beam and through advanced machine
vision analysis.
thickness graph across the glass ribbon and
Data
optional edge distortion in milliradians and a
output:
colour-coded thickness history of the glass
ribbon oven speed and ribbon width variation.
Measuring 3.8 m.
width:
Units:
thickness in mm.
Accuracy: 0.01 mm.
standard from 1 to 6 mm. Custom design is
Glass
thickness: also possible
Scanner computer-controlled scanning across the
glass surface.
~

Fig. 2. On-line glass thickness measurement.

5. Conclusion
Many great innovations never reach their markets due
to shortage of resources to develop those ideas to
industrial products, to get required awareness in industry
and to exploit the solutions in global marketplace.
EC supported technology transfer and take-up
programmes have tried to help young technology
companies to go over the chasm between innovation and
early market entry. The impact so far has been successful,
but there is still a lot to do and new ways to disseminate
the achieved results in different trials have to be
developed.

6. References
[I] A. J. Soini, Machine Vision News, proceedings of Machine
Vision Seminar, Finnish Society of Automation, Helsinki,

Finland, 26.-27.5.1998, pp. 7-9.


Machine Vision Online, AIA,
htttx//www.niachinevisiononline,ord
[2]

USA,

2001,

[3] HPCN-TTN initiative, httD:/lwww.hmn-ttn.org

[4] Integrated Machine Vision Cluster, EUTIST-IMV,


htto:/lwww.eocc.ed.ac.uk/ttn/eutist/

337

[5] A. J. Soini, "New sensors for new machine vision


applications" Sensor Review, MCB University Press, Bradford,
England, Volume 20, No 4/2000.
[6] A. J. Soini, "Technology transfer from research to industry"
The 1 1 l h Scandinavian Conference on Image Analysis, SCIA'99,
Kangerlussuag, Greenland,, June,7-1 1, 1999, pp 35 1-357.

[7] JaaMto Vuorilehto, "The benefits of continuous dimension


control in sawmills" Machine Vision News, Vision Club of
Finland, Helsinki, Finland, Volume 4,1999.
[SI K-E. Peiponen, "The optical inspection of float glass"
Machine Vision News, Vision Club of Finland, Helsinki, Finland,

Volume 4, 1999,a web version is available


httv:liwww.automaatioseura. fiiiaostotimvn/.

in:

338

You might also like