International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
this goal. In this regard the communication between
the computer system and production lines is of crucial
importance, since the collection of visual features is
fed to the computer system for processing, analysis
and decision making, the decision made is ought to be
fed back to the production lines.
4. The frame generating rate: Which is one of the
characteristics of these systems rather than one of
their components. The frame generating rate in these
systems ranges from few frames to few thousands
frames per second. As a general rule, the higher the
frame rate is the better, but raising the frame rate in
many cases happens on the „cost” of the resolution,
so choosing a system that balance the resolution and
the frame rate suiting both of them to the application
at hand is of at most importance.
1.2 Embedded closed systems
These systems are known by their popular name, the "smart
cameras". In these smart camera systems a CCD or CMOS
sensors coupled with processor (CPU or RISC) DRAM
memory, and In/Out facilities are incorporated and embedded in
one package making for the smart camera. Because smart
cameras function as complete vision system, and because it is
almost always impossible to modify the build set up of these
cameras, so their individual characteristics have to be evaluated
fully and thoroughly when the task of implementing such
cameras comes to one's hands. The determinant characteristics
of smart camera that ought to go through study and evaluation
process before choosing a smart camera for an inspection task
at hand are the following:
l. Sensor resolution: Sensors implemented in smart
cameras usually one of two kinds, either CCD or
CMOS, although CCD sensors seem to be more
common in the smart cameras. The resolution of these
sensors is still in the range of submegapixel to few
megapixels, sometimes they exceed this resolution
but this generally happens on the expense of spectrum
richness.
Embedded Processor and memory: The processing
requirements in smart cameras are carried out by an
embedded processor rather than a separate computer.
The most common processor implemented in smart
cameras is the Motorola Power PC, it is not
uncommon to find the Intel Processors built in these
cameras too. Weather it were Power PC or Ix86
Processors these provide the designers of these smart
cameras with ample tools of computing and control
abilities. Sometimes these embedded processors are
supported by task — specific facilities to boost their
performance like floating point processors, digital
signal processors, and the like. Beside the processor
the smart cameras are usually provided with 16 — 32
MByte DRAM memory to provide the processor (and
the camera) with fast processing and storage facility.
3. Application Programming Interface (API): These are
software packages, mostly written in C/C++
programming language, and developed mainly to
perform machine vision tasks, like extracting useful
data from captured images, processing the extracted
data, and generating reaction data (parameters)
accordingly. If the machine vision tasks are well
defined, than the application programming interface
(API) program can be satisfactorily designed and
developed enabling the smart camera to provide an
DD
>
478
effective alternative for PC based open systems.
However, once the API package is designed and
developed, it will be difficult if not impossible to
carry any change or modification on the package, a
property that turns smart cameras to a nonflexible
machines.
4. Programmable 1/0 Facilities: — Programmable
Input/Output facilities are an important tool that
enables setting the smart camera parameters to suite
achieving the task of machine vision at hand. The
programmable 1/0 plays a mediatory role between
conditions present at production environment and
reference conditions and actions.
5. Video Output, Networking, and Interfaces: These are
tools to enable access, communication and graphic
data and event feed to/from the camera from/to the
outside surroundings. These tools include but not
limited to configurations like SUGA output, serial
and/or parallel interfaces, digital 1/O lines, Ethernet
connections, Hotlink lines, etc.
In few words, since smart cameras are generally deployed to
perform specific task, they excel when the task required to be
served is very well defined and can be expressed in terms of
well designed series of data extracting and handling steps.
On the contrary of this, if the task at hand can’t be defined in
specific steps, or if the conditions associated with carrying out
the task have the tendency to change in the future, or if the task
to be achieved itself might be modified, then it is better to opt
for an open machine vision system, which makes distinction
between the imaging unit of the machine vision system and the
computer system that guides, controls and handles the whole
machine vision system.
In this study, choosing flexible open-structure machine-vision
system was adopted due to the fact that this study had two main
goals. First, the investigation of the
implementing quality assurance system as part of the
production lines of pita and tortilla breads. Second, developing
procedures for collecting visual features of these products.
These procedures might be generalized and later applied in
more complicated pattern recognition applications.
2. SIZE AND SHAPE MEASUREMENTS
Pita bread and tortillas are similar in shape and colour, the
differences between these two kinds of breads is that pita bread
has two layers like a pouch or pocket while tortilla has one.
Consequently, during backing phase pita bread will get a
spherical shape while tortillas remain flat. However, when pita
bread cools down the upper layer which is the one having
spherical shape collapses down to the lower layer which
remains flat during backing. This collapsing process results in
flattening both layers, this process happens within about five
minutes after the loaves roll out of the backing belt, and it is
most apparent if the loaves are stacked over each other. As a
result any measurement on the shape of pita bread has to take
place immediately after they roll out of backing belt (before
collapsing). The other difference between pita and tortilla
breads is that the dominant colour in pita is lighter brown with
darker brown spots, while tortillas have a light colour as the
dominant colour with darker brown spots.
In relation to visual features measurements, the procedures and
methods of collecting measurements data are the same in both
kinds of breads, so the discussion here will be related only to
pita bread, while it can be generalized to tortillas without any
change. The size and shape measurements carried were the
following:
possibilities of
Inter
2.1
The
view
is us
one
the «
45 c
num
(R) «
If th
loaf,
of |
extei
of th
2.2
The
detei
line
the |
heig
2.3
Loat
sim
disk.
In th
bea
and
area
2.4
Loaf
eque