'anbul 2004
n systems.
signed and
possible to
package, a
nonflexible
ygrammable
t tool that
ers to suite
hand. The
le between
nment and
;: These are
ind graphic
from/to the
de but not
itput, serial
es, Ethernet
deployed to
juired to be
in terms of
steps.
> defined in
carrying out
r if the task
etter to opt
. distinction
tem and the
s the whole
chine-vision
1d two main
ibilities of
art of the
developing
e products.
applied in
TS
colour, the
it pita bread
lla has one.
will get a
r, when pita
one having
ayer which
ss results in
| about five
It, and it is
other. As a
has to take
belt (before
and tortilla
brown with
jlour as the
cedures and
ame in both
ited only to
without any
d were the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
2.1 Diameter and Radius
The pita bread diameter (D) can be measured from an upper
view (or front view) image of the loaf. If an upper view image
is used to measure the diameter, then the number of pixels from
one point on the perimeter of the loaf to its counterpart point on
the other end of the perimeter can be calculated by moving in
45 degrees toward the centre point of the loaf circle. If the
number of pixels calculated is multiplied by the pixel’s radius
(R) of the loaf which is half the diameter (R=D/2), figure 1.
If the diameter (D) is calculated from a front view image of the
loaf, then the diameter will be in this case equal to the number
of pixels along one horizontal line (the diameter itself)
extending between the most left point and the most right point
of the loaf’s perimeter, multiplied by the pixel width.
Figure 1. Arabic pita bread
2.2 Height
The loaf height is the highest point in the loaf. It can be
determined by calculating the number of pixels along a vertical
line extending from the peak point of the loaf and base line of
the loaf and multiplying the number of pixels by the pixel’s
height, figure 1.
2.3 Loaf area
Loaf area can be calculated in one of two ways. The first one
simple calculation in which the loaf is considered to be a flat
disk, the area simply calculated as:
$ = 1x R (1)
simple
In the second and more accurate case, the loaf is considered to
be a sphere sector; this case is true right after the baking phase
and before the collapse of the loaf. Here the loafs surface (face)
area is calculated as:
S, zzx(R 4 H^) (2)
up
2.4 Loaf volume
Loaf crust volume is calculated according to the following
equation:
V-zxHx(xR - H^)/6 (3)
479
3. VISUAL FEATURES EXTRACTION
3.1 Colour Features Measurements
At the beginning and during the backing phase, all conditions
are set to produce pita bread loaf with ideal characters.
Conditions such as dough's water content (humidity), thickness
of loaf dough, movement speed of backing belts, and backing
heat are all important factors in producing ideal crusty brownish
pita bread.
In many cases, however, the resultant dominant backing colour
of the bread is not always the same as the wanted backing
colour. The ideal backing colour is determined by choosing
samples from areas that were successfully backed to the wanted
brownish colour, then, the average value of the grey level of
these samples is determined and adopted as the grey level
representing the ideal backing colour.
Dominant colour is defined as a spectrum range with minimum
and maximum grey level values centred around the grey level
value associated with the greatest number of pixel found in the
pita bread image. The two maximum/minimum thresholds can
be adjusted at will; this will result in significant flexibility in
determining the dominant colour by including/excluding wider
range of the colour spectrum (grey level). Light and dark areas
in the pita bread loaf are those areas exposed to less or
excessive backing heat. In the context of this study they are
defined as the summation of lighter/darker pixels deviated from
the ideal backing threshold.
The average colour of the whole pita bread's loaf is an
important factor in determining the success in obtaining the
required backing colour or approximating that goal. The
average colour of the whole pita bread's loaf (Average Grey
Scale: AGS) is determined by computing the summation of the
grey levels of all pixels (Grey Scale of Pixels: GSP) and
dividing it by the number of pixels in the loaf (Number of All
Pixels: NAP). The following equation illustrates the calculation
of the average grey level of the loaf.
D> GSP
AGS = ——
NAP
(4)
3.2 Image Processing Procedures
Visual feature extraction was carried out on the pita bread
images using a combination of program code developed
specifically for this project and the commercially available
image processing extension package of the MathCAD 2000
program. Image processing procedures that were implemented
in this study are the followings:
3.2.1. Reading Image Files: Images of the incoming loafs of
bread are captured in the JPEG image file format. Then the
binary data is converted to a matrix carrying the values of the
grey scale of the pixels as it its elements, and each element in
this matrix will be associated with row and column coordinates.
If the image captured is saved as a row monochrome image,
then the image can simply be converted to an (M*N) element
matrix, where M is the image's rows, and N is the image's
columns. On the other hand if the image captured is saved as
RGB coloured image, then the image can be converted to an (M
x 3N) element matrix, where M is the image's rows, and 3*N is
the image's columns. This apparent triplication of the columns