Industrial robotics and automation .: Makina Mühendisi

Industrial robotics and automation

This section examines the ‘robotic’ equipment used in manufacturing industry and includes reprogrammable automation
equipment. NC machine tools and automated assembly equipment are excluded except where industrial robots are involvedl.
Similarly, ‘hard’ automation (e.g. cam-controlled mass-production equipment) is omitted. Thus the main subjects
of interest will be industrial robotics and artificial vision.

Industrial robotics

Definition

It is important, in an industrial context, to have an unambiguous definition of a robot. All the major industrialized
countries have adopted definitions with some more complex than others. The British Robot Association definition is
succinct and is therefore selected for use here, i.e. ‘An industrial robot is a reprogrammable device designed to both
manipnlate and transport parts, tools or specialized manufacturing implements through variable programmed motions for the performance of specific manufacturing tasks’. Industrial robots are therefore reprogrammable manipulators of tools, materials or components. A typical robot is shown in Figure 5.11. They should be easily reprogrammed to carry out work on new tasks and they should have a degree of dexterity. Thus an NC machine tool is not a robot since, although it is easily reprogrammable, it is not designed to do anything other than cut material. Neither is the type of arm used to manipulate radioactive material in the nuclear industry; these devices are constantly under the control of a human operator and are therefore remotely controlled, Le. they are not programmed to operate autonomously.

Reasons for using industrial robots

In modern manufacturing there are many advantages in using robots rather than human labour or hard automation. In
relation to human labour:

Robots work to a constant level of quality. Waste, scrap and rework is minimized. They can work in areas that are hazardous or unpleasant to humans.

No jobs are boring, tiring or stressful to robots. Continuous 24-hour production is possible for many days.
They are a single investment; salaries do not have to be paid each year at increasing rates, and there are no burden
costs such as pension and insurance schemes, holidays, sick pay, etc. Investment in a robot involves a once-only capital expenditure, whereas human labour requires an ongoing salary cost that increases annually. Robots are advantageous where strength is required, and in many applications they are also faster than humans.

Also, in relation to special-purpose dedicated equipment, robots are more easily reprogrammed to cope with new
products or changes in the design of existing ones. Dedicated equipment usually requires expensive strip-down and rebuild
in these situations and often has to be discarded as obsolete. As well as these obvious reasons there are other, indirect,
advantages to be gained from robotization:

0 When changing from manual to robotic methods, the product components will often have to be redesigned to
provide simplicity of presentation, positive gripping points, unambiguous orientation and location, adoption of the stacking principle for assembly, and ease of location for screws, etc. This usually results in a simplified, better and cheaper design for the product.

Quality will be improved in many areas as automatic inspection techniques are adopted. Design changes can be implemented more quickly and new products introduced efficiently. Lead times can be reduced. Work in progress can be reduced. In comparison to dedicated equipment, smaller batch sizes can be handled and downtime between product changeovers is reduced.

It should be remembered that robots are simply one alternative and that human labour and dedicated special-purpose
equipment also have their place in the manufacturing environment. Although much simplified, Figure 5.12 shows the
cost-to-volume region in which robots are most attractively employed. Generally, unless there are severe environmental
or hazardous conditions, human labour is suitable for lowvolume high-variety work. Conversely, for very-high-volume/
low-variety work dedicated equipment, or ‘hard’ automation, is probably the most cost-effective.

The construction of industrial robots

Essentially an industrial robot consists of two elements - the manipulator (or ‘arm’) and the robot controller. The controller
contains the microprocessor system and the power control units. Hydraulic and pneumatic robots also have pump
and compressor units, respectively. The particular geometry of the arm will provide an associated work envelope; the arm
will be powered by electrical, hydraulic or pneumatic means; it will be non-servo or servo controlled; it will be programmed
on-line, off-line or both; and will be capable of point-to-point, point-to-point with coordinated path, or continuous path
movement. The following is a brief explanation of each of these terms.

The control unit This unit interfaces with the robot’s internal and external sensors, drive units, peripheral equipment, and
the programmer and operator. It is therefore usually capable of handling serial and parallel data transmission at various
rates, and can carry out digital-to-analogue and analogue-todigital conversion as necessary. Communication with the
programmer is via a visual display unit and keyboard, or a teach pendant. There will also be floppy disk drives for
loading and saving programs and a printer for hard copy.

Program interpretation will be carried out within the controller.

Some robots are capable of using more than one language. In this case the language and operating system is
usually loaded from disk at the beginning of the programming session. Within the unit there will be ample memory space to
store all necessary data for coordinate transformation, trajectory computation, monitoring and decision making.
The unit will also perform the functions necessary for full servo control. In some robots each axis has its own microprocessor
system supervised by a ‘master’ system. There may also be a system dedicated to handling the sensory data input. Thus
in a six-axis robot there may be eight integrated microprocessor systems within the control unit.

Control of the power units is effected by the control unit. In electric-drive robots, low-voltage control signals are sent out
to the motor drive amplifiers, one for each axis, to produce power for the motors. Servo control is maintained by monitoring
feedback from internal sensors. In fluid power systems, solenoids and servo-valves are also controlled in this way.
Robot geometry The robot arm is composed of links and joints. The joints, also referred to as articulations or kinematic
pairs, will normally each have only one degree of freedom. In robots this means that a joint will probably be (1) revolute (i.e.
rotation about one axis), (2) prismatic (i.e. linear movement along one axis) or (3) screw (i.e. a combination of linear and
rotational movement along one axis with translation defined by the screw pitch and the rotational displacement). Other
types of joint with more than one degree of freedom (e.g. a ball joint) are more difficult to control.
The number and relationship of these link and joint arrangements defines the dexterity of the robot arm. There is a
maximum number of six degrees of freedom available to any free body in space. For many industrial applications (e.g. arc
welding, fettling and spray painting) six degrees of freedom are desirable. With a minimum of six joints it is possible to
achieve this. However, the design of the robot must be carefully considered as there will be ‘no-go’ regions into which
the robot arm will not be able to reach due to its physical limitations. Also, even though a robot may have a large
number of joints, it may not have the equivalent number of degrees of freedom. For example, an arm with six revolute
joints whose axes are all parallel will only have three degrees of freedom.

Usually a robot arm has three major axes providing three degrees of freedom. These axes allow the robot to position its
end effector or gripper at any point in space within its work envelope. In addition to the major axes there will also be one,
two or three additional axes, normally in the form of a ‘wrist’ at the extremity of the arm. These will allow the robot to
orientate its end effector, which will be fixed to the wrist, about any point in space. As mentioned previously, the
number of degrees of freedom required depends on the task to be performed. For example, population of a printed circuit
board will only require four degrees of freedom. This operation is essentially a pick-and-place type, with one vertical, two
horizontal and one rotational movement for component orientation.

However, for arc welding a complex three-dimensional seam (say, at the intersection of two cylinders) a robot with
the full six degrees of freedom wil be necessary. This is to ensure that the welding gun is constantly maintained at the
correct orientation to the work as welding proceeds. Indeed, this particular task may demand additional axes to be
employed by clamping the work on a multi-axis servo-motor controlled worktable whose movements wil be integrated with
those of the robot and controlled by the same control unit. Some typical robot configurations and work envelopes are
shown in Figure 5.13.

Robot drive and control systems Many of the first industrial robots were hydraulically driven. However, most robots now
produced are electrically powered. The previous definition of a robot implies relatively sophisticated control, and this is
verified by the fact that nearly all industrial-quality robots are fully servo controlled. Non-servo controlled robots, just on the
borderline of the definition between robots and simple pickand- place units, are usually driven pneumatically. Some very
light-duty devices, and often the orientation axes on SCARA robots, use non-servoed stepping motors.
Electric robots are usually driven by d.c. permanent magnet servo motors, brushless motors or, occasionally, stepping
motors. Electric-drive systems are relatively clean and quiet when compared to fluid power machines. They are easily
maintained and repaired and are well suited to electronic control. Recent developments in the use of rare-earth materials
for permanent magnets mean that power-to-size ratios are increasing, and the use of brushless motors reduces
maintenance costs. Brushless motor drives can also be used in areas such as clean rooms, since contamination particles are
reduced, and in situations where there would previously have been a fire risk due to the possibility of brush arcing. Unless
incorporating direct-drive motors, electrically driven robots do have the disadvantage of requiring transmission systems.

These add cost and weight, and also reduce precision due to gear backlash and other unwanted movement.
Hydraulically powered robots still have some advantages.

For example, they have very good power-to-size ratios and hydraulic force can be applied directly at the desired point
without the need of a transmission system. Hydraulic fluid is incompressible and therefore there are no backlash problems.
Assuming the power pack, which contains the electrically driven hydraulic pump, is located remotely, then the robot can
be used in high fire-risk areas. This is because only very low voltages for control and feedback purposes will be present on
the actual robot arm. Because of their necessarily sturdy construction due to the high hydraulic pressures experienced,
they can withstand higher shock loads than other robot types. However, it is their disadvantages that have led to their
reduction in popularity. A noisy power pack, even when protected by an acoustic muffler, makes them environmentally
unattractive. Historically, they have tended to be less reliable than electric robots with leakages occurring which contaminate
work areas and cause performance loss. Servo control of hydraulics is not as simple as that for electrics and availability
of skilled personnel is more scarce. The viscosity of the hydraulics fluid can be affected by temperature and this can
cause variations in performance. Finally, cost is not directly proportional to size. Smaller hydraulic robots tend to be much
more expensive than their electric counterparts.

Pneumatic powered robots are the cheapest and least sophisticated type. They are usually not servo controlled but will
be able to carry out complex movement sequences if necessary. They are fast, simple, reliable and easily understood by
most factory technicians and maintenance personnel. They also have the advantage of being intrinsically safe and can
therefore be used in explosive atmospheres. The major disadvantage of pneumatic robots is that precise servo control is not
practical. This is due to the compressibility of air, particularly when moving heavier loads. Thus pneumatic robots are
usually found in limited-sequence. light-load fixed-speed applications.

Path control

The application will influence the choice of the robot path control system. Robots with simple point-to-point control are
suitable for assembly, palletizing and other materials-handling

Robot classification by geometric configuration and work envelope. tasks. Point-to-point with coordinated path control is suitable
for tasks such as arc welding, sealant application and spot welding of moving components (e.g. car bodies on a conveyor
system). Continuous path control is used where the dexterous movement of a human operator has to be mimicked (e.g. in
spray painting, or where complex contouring movements are necessary).

In point-to-point control the robot will move between defined points without regard to the path taken between them.
In some robots an additional software facility allows the choice of movement between points in either the shortest travel time
or in a straight line. In revolute robots movement in an ‘elbow up’ or ‘elbow down’ mode can also be selected. When using
robots with simple PTP control particular care has to be exercised to ensure that collisions with obstacles will be
avoided when the robot is running.

In point-to-point with coordinated path the control software allows the path the end effector will follow between points to
be determined. Straight lines, circles, arcs and other curves can be defined. Two points are all that is necessary to define a
line; circular movement can be programmed by specifying three points on a circumference or a centre point and a radius.
Full continuous path control is most often obtained by playing back the information recorded from physically leading
the robot through a desired task. Every movement of the arm is recorded in real time by sampling joint positions at a high
frequency, and this is used in on-line programming, Continuous path control is also employed in off-line programming in
some systems where there exists the facility to insert the mathematical equations for desired curves. These curves will
then be followed by the end effector.

Robot programming methods

Industrial robots may be programmed using a number of techniques. The most basic methods, employed in some very
early hydraulic robots, used rotating drums on which pegs could be set to close microswitches. These switches operated
solenoid valves, so controlling the flow of fluid to the actuators.

Adjustable mechanical stops were fixed on the moving members to contact limit switches at appropriate points.
These mechanical programming methods provided control of sequence and distance moved. More recently, the use of
programmable logic controllers has been applied to sequence the movement of pneumatic robots and modular units. These
are all non-servo control systems employing simple feedback from limit switches or proximity sensors.
Full servo controlled robots employ dedicated microprocessor- based controller units as described earlier. This
allows sophisticated on- and off-line programming techniques to be employed. On-line programming Here the robot arm itself is used during the direct programming operation. This method has the following advantages.

The robot can be observed as programming progresses; this increases confidence in the finished program since
possible collisions and other robot ‘no-go’ areas will be easily identified. In applications such as spray painting and welding where
human experience is important, the direct programming of the arm by an expert effectively produces a transfer of
skill from the human to the robot.

This type of programming is easy to learn and can often be accomplished by the personnel that the robot is replacing.
Less expensive computing hardware and software is involved than in on-line programming,

On-line programming also has one main disadvantage. Where the task being programmed is complex the programming time
may be prohibitively long. For example, if a deburring operation is to be carried out and the robot has to be taught the
shape of every hole and curve. the programmer will find the exercise extremely tedious and the cost benefits may be trivial.
It is therefore a technique suited to tasks which are highly repetitive in nature, Le. one sequence of movements can be
repeated automatically a number of times.

There are basically two methods of on-line programming: teach by lead-through and teach by pendant. It is also possible
to program directly from a computer terminal attached to the controller, which is similar in principle to the teach by pendant
method.

Off-line programming Off-line programming involves creating the program for a robot task, without the need to be
connected physically to the robot or even to be anywhere near its physical presence. In fact, when coupled with simulation
techniques, off-line programming can be carried out before deciding on which robot to purchase for a specific application.
Some of the advantages of off-line programming are as follows:

1. The robot for which the program is being made can continue working on its old task until ready for the new
program. This obviously reduces robot down-time and increases productivity.
2.If the control system and language allows, it is possible to build into the program collision avoidance, error recovery
and other contingency routines.
3.Compared to on-line programming, it is easier to make alterations to cope with variations in products and design
changes.
4.Off-line programming is suited to full computer integration of a facility. For example, if a robotized computercontrolled
machining cell is in operation, then with offline programming the problems of downloading programmes to the robot at the appropriate times is greatly reduced.

There are also some disadvantages:

1. Real-world contact is lost.
2.More expensive hardware and software is required.
3.More programming skill is needed.
4.Fine adjustments under production conditions on the
shopfloor are usually necessary.

Sirnulation Graphical simulation of the robot and its environment offers many benefits to the industrial engineer and
programmer. Figure 5.14 shows an example of a low-cost simulation system display. This particular software, called
‘Workspace’, runs on a PC. Simulation packages provide the ability to model the robot kinematically and often dynamically
to give an animated, real-time, visual representation of how the robot will work under programmed conditions. Threedfimensional
wire frame - sometimes with colour shading or solid modelling - techniques are used.

There are a number of advantages associated with simulation:

1. Simulation allows a prospective robot purchaser to try out various models at relatively little cost before making a
decision. Parameters such as work envelope, cycle times and joint configuration limitations can be compared for all
the robots held in the library.

The immediate robot work area can be simulated and various permutations of machines and operation sequences
experimented with before finalizing the layout.

2.The immediate robot work area can be simulated and various permutations of machines and operation sequences
experimented with before finalizing the layout.

3.Potential collisions can be detected at an early stage and programs and layouts modified to suit.
4.Simulation is very suitable for teaching and training purposes. Mistakes in programming can be observed and
learned from without the hazards and potential costly damage that would be experienced in the real world.
5.At the robot design stage, simulation delays the need for physical prototypes to be built, thus reducing research and
development costs.
These advantages, coupled with improving computing powerto- cost ratios, mean that simulation is becoming an increasingly
popular robot-programming tool in industry, education and research.

Using industrial robots

Suitable applications The potential advantages of robotization can be maximized by making wise application selections.

Industrial robots realize their full economic potential in applications where product volume is large enough to recoup the
expenditure on hardware, programming and engineering costs, yet is sufficiently low to prevent justification of dedicated
special-purpose equipment. However, high-volume work with frequent model or option changes, such as is found
in autombile assembly, is suitable for robotization. The effect of robots on the cost per unit, in relation to volume, is shown
in Figure 5.15. The following are some further indicators as to applications that should provide suitable opportunities:

1. Tasks which are carried out in (or create) an unpleasant or hazardous environment. For example, toxic or flammable
atmospheres are created by processes such as arc welding and spray painting, and removing human operators from
these jobs can improve quality and increase production rates.

2. Jobs that are tiring or boring. Robotization of these eliminates absenteeism and labour turnover problems and
usually improves quality.

1.Repetitive and simple operations requiring simple movements allow the least expensive robots to be used and
minimize installation and programming problems.

2.Desired cycle times should not be too short. For example, if the cycle time is greater than, say, 3 seconds then the
choice of robot is relatively wide. However, if very short cycle times are required, as in PCB component placement,
then more specialized and expensive high-speed robots will be necessary.

3.The tolerances on the components and tools should allow robots of average precision to tackle the work.

4.The variety of products expected to be handled by the robot should not be large nor changes from one product to
another too frequent. This will keep engineering and reprogramming costs to a minimum.

5.The following points are also relevant. If the task being considered has an integral inspection element, then additional
costs will be incurred when vision or other sensing methods are added to carry out that inspection.

In materials-handling applications very heavy loads will demand larger and more expensive robots. If an ordered environment exists (or can be made to exist) around the robot then robotization is simplified. If possible, work should be oriented and positioned at the previous operation before presentation to the robot. Most robots available commercially have limited reasoning ability, therefore tasks should demand little in the way of intelligence or judgement.

Once the task to be robotized has been selected the next stage is the selection of an appropriate robot.
Selecting the robot Robot selection should be carried out after listing task demands such as cycle time, payload required,
necessary precision, and cost. These demands will be compared against the specifications provided by the robot
supplier or manufacturer, some of which are now listed.

1.Speed. Having decided the speed required of the robot from the work analysis, the detailed specification should
now be examined. Some manufacturers may give maximum speeds for each axis of the robot, some the maximum
speed of the end effector. These should be given for maximum load and at maximum reach as well as for under
optimal conditions. It should be remembered, however, that maximum speed is not necessarily a very useful piece
of information, because a robot arm must accelerate to and decelerate from this speed. Figure 5.16 shows a
typical robot velocity curve. For some applications, particularly assembly, a 'goalpost' time is a more useful
specification. This is the time, supplied by the robot manufacturer, that it should take the robot to complete a
standard series of movements carrying a standard load. For example, the movement may be close gripper, move
up 30 mm, move across 300 mm. move down 30 mm, open gripper. This will prove more useful for estimating
than a maximum speed figure.

2.Payload. The maximum load expected to be encountered will have been determined, and a robot with sufficient
strength to handle a considerably greater load should be selected. The specification should show whether the maximum
load capacity is given with the arm close to the body or at full extension where the capacity will be much less
due to leverage. Robots are available with capacities ranging from a few grams to 2 tonnes.

3.Precision. The overall precision of a robot is composed of three elements, i.e. resolution, repeatability, and accuracy.
The resolution of a robot is normally a feature that is transparent to the user and is therefore not included in
standard specification sheets. It refers to the smallest

controlled movement the end effector is capable of making. This is determined by (a) the resolution of the
computer controller (i.e. the number of bits used to define a position over a given range), (b) the resolution of the
drive system (e.g. the number of steps per revolution provided by a stepper motor and associated gearing) and
(c) the resolution of the feedback elements such as shaft encoders. The repeatability of a robot is determined by its
resolution plus clearances and wear on moving parts plus any other inaccuracies and errors in the total system. It is
a statistical term describing how well the robot can return consistently to a taught point. This is the most common
figure relating to precision to be included in robot specification sheets. Repeatabilities of tl or 2 mm mediumduty
work. f0.05 to f0.03 mm for medium assembly, and t O . O 1 mm for precision assembly are typical.

3.Accuracy. Assume that a computer-controlled robot is to move to a point in space. This point is defined by entering
the coordinates into the control system. The difference between the taught target point and the actual position
achieved by the robot, in the real world, is the 'accuracy'. 4.This will be determined by the resolution, inaccuracies in
the 'model' of the robot held in memory and other factors such as bending or thermal expansion of the robot arm.
The relationship between accuracy and repeatability is shown in Figure 5.17.

5.Configuration. The supplier will provide information on the geometric configuration and dimensions for the effective
work envelope of the robot. These can then be used to construct templates either on card or on computer to
enable an appropriate work layout to be designed.

6.Control system and programming method. The specification will provide information on whether point-to-point,
point-to-point with coordinated path or continuous path control is provided. It will also state the programming
methods used. For CP programming by lead-through a slave arm may be available and for PTP or PTP-CP teach
by pendant methods may be used. For many robots programming using a computer terminal and a high-level
language will also be available.

7.Cost. The cost of a complete robot installation can vary considerably from that of the basic robot. The robot
chosen can influence this total cost. Ease of programming, and interfacing capabilities will influence the engineering
costs. Cost of fixturing, parts presentation and orientation devices, and end-of-arm tooling will have to be included in
the total. Also, if working to a fixed budget for the robot, there will probably have to be a trade-off between precision,
speed, strength and reach.

Qther specifications that should be considered include drive system, number of degrees of freedom, type and number of
input and output ports. and memory size.

Robot safety As well as presenting the normal safety problems associated with moving equipment that is electrically,
hydraulically or pneLmatically powered, and machines that are under microprocessor control, industrial robots present some additional problems that are unique:

1. While executing a program the robot appears to the inexperienced observer to be moving spontaneously and
unpredictably, each movement being difficult to anticipate. This applies particularly when the robot is at a 'dwell' point in its work cycle. It may appear to be deactivated but in fact it will spring into action as soon as it receives an appropriate command from the system controller.

2.Most robot arms sweep out a work volume much larger than that occupied by their base. With a six-degree-offreedom
robot the arm movements and positions are therefore difficult to visualize.

3.Heavy-duty robots are built to be rugged and inelastic, and fast-moving arms are therefore extremely dangerous.
Size is not necessarily important as was proved when one person was killed when struck on the back of the neck by a
small teaching robot. Accidents can be caused by human

4.The integrity of the control system hardwarez and software is particularly crucial since faults will produce erratic and unpredictable behaviour.

carelessness, insufficient training, poor robot or installation design, poor quality components used in the system,
and software errors. Most accidents occur to those familiar with the robot such as programmers, maintenance
engineers and operators. Those unfamiliar with robots tend to be more wary - it is the complacency caused by
familiarity that is dangerous.

Industrial robot safety should be considered at the stages of robot design, supply, installation, programming and everyday
usage. The designer should ensure all controls conform to good ergonomic practice. Controls and displays should obey
standard conventions, mushroom-shaped stop buttons should protrude from surfaces and there should only be one start
button, which should be recessed. All emergency stops should be hardwired into the power supply and not rely on software
execution. Conventional good design practice should be observed, moving parts should not be exposed and there should
be no trapping points for limbs or fingers, with no unnecessary protrusions capable of inflicting injury.
The robot supplier should ensure that proper instruction and training is given to appropriate personnel designated by
the purchaser. The supplier should also make the user fully aware of the robot’s limitations and any possible hazards that
may be encountered.

When the user is planning and implementing the installation full consideration should be given to the robot’s position
within the factory, e.g. it should not be located near any trapping points such as roof pillars or stanchions, and it should
not be possible for it to reach into passageways or manual work areas. Preparation of safety manuals, or safe working
procedure documentation may also be carried out at this stage. Reference should always be made to appropriate rules,
regulations and guidelines. In the UK there is the Health and Safety Executive (HSE) guidance booklet Industrial Robot
Safety, and the MTTA booklets Safeguarding Industrial Robots Parts 1 and 2. There is also the British Standard 5304
Code of Practice - Safeguarding of Machinery which contains the basic principle of safeguarding, i.e. unless a danger point
or area is safe by virtue of its position, the machinery should be provided with an appropriate safeguard which eliminates or
reduces danger. Light curtains and pressure-sensitive mats are commonly used around the immediate vicinity of the robot.
For maximum safety a 2 m high cage around the robot is recommended. This should have doors electrically interlocked
to the power supply to ensure that unauthorized entry deactivates the robot.
During everyday operation management must ensure that only fully trained personnel operate the robot. Established
safety procedures must be adhered to, appropriate warning signs given high visibility and, generally, a state of continual
safety ‘awareness’ cultivated.

Industrial vision systems
Vision system components

The basic elements of an industrial vision system are shown in Figure 5.18. A camera is first necessary to acquire an image.
This camera may be a vacuum-tube or a solid-state type, the latter now being the most popular. The signal from the camera
is then processed in the vision system computer. The image observed by the camera and the digitized image used for
computer processing are observed on a monitor which is switchable between them, or two monitors may be used, one
for each image. A means of communicating with the system is necessary and this would take the form of a computer terminal
and visual display unit. A means of allowing an automated physical reaction in response to the vision analysis is required.

This demands an interfacing unit connected to a robot or other device (say, a simple pneumatic cylinder for rejecting bad
parts). Finally, to ensure optimum viewing conditions special lighting arrangements may be necessary to avoid distracting
shadows or glare.

Vision system types and operation
Vacuum-tube cameras provide an analogue voltage proportional to the light intensity falling on a photoconductive target
electrode. This electrode is scanned by an electron beam and the resulting signal is sampled periodically to obtain a series of
discrete time analogue signals. These signals are then used to obtain digital approximations suitable for further processing.
For example, if an A/D converter has a sampling capability of 100 ns, the image is scanned at 25 frames per second and each
scan is composed of 625 lines, then there will be 640 picture elements, or ‘pixels’, per line.* Since some time is lost as the
electron beam switches off when moving from one line to the next, the number of pixels in a frame will be 625 X 625. which
gives a total of almost 400 000 pixels.

This is difficult and expensive to handle computationally in real time, especially if mathematical analysis of the image is to
be carried out. For this reason, the number of pixels can be reduced, depending on the application, to provide a more
manageable image.

Solid-state cameras use arrays of photosensitive elements mounted on integrated circuits. The light from the scene is
focused by the camera lens onto the IC chip. Charge-coupled devices (CCDs) or photodiode arrays are scanned to provide a
voltage signal from each light-sensitive element. These solidstate arrays are available in many densities, often from
32 X 32 to 1000 X 1000 pixel arrays. The larger pixel densities are too high for real-time vision analysis but they do provide
high-quality video pictures. The voltage signals from the photosites are again digitized before further processing. In a ‘line
scan’ camera linear arrays of photosites are used rather than the area type. Line scan cameras can be used where the object
is moving steadily across the field of view. For example, an object passing under the camera on a conveyor belt can be
scanned repeatedly and an image built up in the vision system memory.

Recently, solid-state cameras have become widely available and relatively inexpensive due to their large-volume production.
They have a number of advantages over the vacuum-tube

* Pixel is derived from ‘picture element’ and usually represents the smallest
possible display feature or point on the display device.

type, i.e. they are much smaller and lighter, more robust, more reliable, use less power, have a broader temperature
operating range at lower temperatures, and are less likely to be damaged by high light intensities.
The pixel voltage signals from the camera are now assigned to a finite number of defined amplitude levels. The number of
these ‘quantization’ levels is the number of ‘grey levels’ used by the system. An 8-bit converter will allow 256 grey levels to
be defined. In practice, this number of levels is often unnecessary and processing time can be reduced by using only 16
grey levels. In some cases only two grey levels, i.e. black and white, are necessary, this is termed ‘binary’ vision.
Each grey level is next ‘encoded’, i.e. it is given a digital code, and the data stored in memory. In a computer vision
system this is done for one picture ‘frame’ and the data stored in memory in what is termed a frame buffer or picture or
frame store. Various algorithms are then used to minimize the data for analysis and organize them in such a way as to allow
feature extraction andl object recognition. Objects are usually recognized by the system by first showing it a sample of the
object. The system ‘remembers’ the object by storing information on features such as object area, perimeter length. number
of holes. and minimurn and maximum radii from the centre of gravity. The sequence of all these processes is shown in Figure
5.19.

Vision system applications

As the capabilities of vision systems increase so also does their popularity. Custom-designed hardware and developments in
algorithms mean that the systems are becoming faster and more reliable. They are now found in a wide variety of
industrial applications and are sometimes supplied as integral components or programmable electronic component placement
machines and other robotic systems. Three main areas of application are listed below.
Identification Here the system is used to identify a product or individual component. For example, it may involve character
recognition, as in reading alphanumeric data on a product label or recognizing a component on a workbench prior to
assembly.

Inspection This is one of the major applications of vision systems as it is estimated that visual inspection accounts for
around 10% of total manufacturing labour costs. This percentage can be very much higher in some industries (e.g. electronic
product manufacturing such as PCBs, computers and other consumer goods). Sensible application choices can prove very
cost-effective. Inspection is generally further divided into ‘qualitive’ and ‘quantitive’ inspection. In quaiitive inspection it
is attributes that are examined (e.g. glass bottles may be checked for flaws or castings checked for cracks, or the
number of pins on an IC chip verified). In quantitive inspection dimensional or geometric features of a product are
measured and checked (e.g. the diameter of a component turned on a lathe or the width of a steel strip coming from a
roliing mill),

Decision making This is a general term which implies a number of applications. For example, a vision system could be
used to guide the welding head of a robotic welder along the seam of a fabrication, or it could assist an automatic guided
vehicle find its way around a factory shopfloor. In conjunction with artificial intelligence techniques vision can be used to
provide the information input necessary to provide autonomous working of rotobic devices in unstructured environments.
In conclusion, vision system technology can be said to be rapidly improving and, in conjunction with advances in related
technologies, it will continue to make a significant impact on factory automation for many years to come.