Chapter 4 Preliminary Explorations
4.1.1 Experiment and Analysis
In machine vision and image processing, Otsu’s thresholding method
Section 2.1.1) is used to automatically perform histogram shape
thresholding. The histogram method assumes tha
foreground and background pixels. In this case, the foreground is the orange, and
the background is the convey
section for testing Otsu’s method.
1. Example One: Testing Otsu
Figure 29 shows a ripe orange
Fig. 29.
Figure 30 shows three isolated colour channels in
Fig. 30
Preliminary Explorations
Intelligent image processing techniques and fruit grading algorithms are modified,
implemented and tested in this chapter. The experiment and analysis are presented
and strengths of previous approaches are explored in order to
a novel algorithm.
tsu’s Method
Experiment and Analysis
In machine vision and image processing, Otsu’s thresholding method
is used to automatically perform histogram shape
thresholding. The histogram method assumes that there is an average value for the
foreground and background pixels. In this case, the foreground is the orange, and
background is the conveying system. Four Examples are presented
s method.
: Testing Otsu’s method on ripe oranges
shows a ripe orange image with no blemishes.
. Ripe orange sample for testing Otsu’s method. shows three isolated colour channels in natural colours.
0. Isolated colour channels for a ripe orange sample.
Intelligent image processing techniques and fruit grading algorithms are modified,
and analysis are presented
are explored in order to
In machine vision and image processing, Otsu’s thresholding method (described in
is used to automatically perform histogram shape-based image
average value for the
foreground and background pixels. In this case, the foreground is the orange, and
presented in this
The threshold values are computed
method.
• Red channel: 96
• Green channel:
• Blue channel: 9
Figure 31 shows the processed images after applying the
channels separately.
Fig. 31. Example of Otsu’s method works perfectly on
average intensity on the blue channel
background pixels appear on t
2. Example Two: Testing Otsu
Figure 32 shows a ripe orange
Fig. 32. Blemished ripe o The threshold values are computed
method.
• Red channel: 94
• Green channel:
• Blue channel: 11
Figure 33 shows the processed images after applying the
channels separately.
values are computed for three channels separately
96
Green channel: 46
9
the processed images after applying the threshold
Example of ripe orange segmentation using Otsu’s method
’s method works perfectly on the isolated red and green channel
blue channel is low due to the nature of the orange
background pixels appear on the right image.
: Testing Otsu’s method on blemished ripe oranges
shows a ripe orange image with blemishes.
Blemished ripe orange sample for testing Otsu’s method.
values are computed for three channels separately
4
Green channel: 53
11
the processed images after applying the threshold
using Otsu’s
thresholds on three
using Otsu’s method.
channel. The
due to the nature of the orange, so some
s method on blemished ripe oranges
using Otsu’s
Fig. 33. Example of Small holes appear on the
on the red channel, such as
performance. In general,
3. Example Three: Testing Otsu
Figure 34 shows an unripe orange
Fig. 34.
Figure 35 shows three isolated colour channels in
Fig. 35
The threshold values are computed
method.
• Red channel: 55
• Green channel:
• Blue channel: 11
Figure 36 shows the processed images after applying the
channels separately.
Fig. 36. Example of
Example of blemished ripe orange segmentation using Otsu’s method.
mall holes appear on the left image which is caused by some low intensity
, such as blemishes and stem. It has no impact on the
general, Otsu’s method works fine for blemished ripe oranges.
: Testing Otsu’s method on unripe oranges
shows an unripe orange image with no blemishes.
Unripe orange sample for testing Otsu’s method
shows three isolated colour channels in natural colours.
5. Isolated colour channels for unripe orange sample values are computed for three channels separately
55
Green channel: 50
11
the processed images after applying the threshold
Example of unripe orange segmentation using Otsu’s method.
s method.
low intensity areas
has no impact on the overall
for blemished ripe oranges.
colours.
using Otsu’s
thresholds on three
Small holes appear on the
reasons. It is evident to see that Otsu
intensity areas on the orange skin
easier to be detected on the red channel. In general, Otsu
unripe oranges.
4. Example Four: Testing Otsu
Figure 37 shows an unripe orange
Fig. 37. Blemished
The threshold values are computed
method.
• Red channel: 67
• Green channel:
• Blue channel: 11
Figure 38 shows the processed images after applying the
channels separately.
Fig. 38. Example of Holes on the left image
channel. Holes on the middle image
green channel, such as blemishes. Holes on the
this case. It is evident to see that the
more important than the others for blemish identification
oles appear on the left and right images which are caused by different
It is evident to see that Otsu’s method is able to identify
on the orange skin, such as blemishes. It seems like the blemish is
easier to be detected on the red channel. In general, Otsu’s method works fine for
: Testing Otsu’s method on blemished unripe oranges
shows an unripe orange image with blemishes.
Blemished unripe orange sample for testing Otsu’s method.
values are computed for three channels separately
67
Green channel: 48
11
the processed images after applying the threshold
Example of blemished unripe orange segmentation using Otsu’s method.
left image are caused by the colour transition areas
middle image are caused by the low intensity areas
, such as blemishes. Holes on the right image are not important in
t is evident to see that the intensity variation on the green channel is
ant than the others for blemish identification.
caused by different
s method is able to identify some low
It seems like the blemish is
s method works fine for
s method on blemished unripe oranges
.
using Otsu’s
thresholds on three
s method.
areas on the red
are caused by the low intensity areas on the
are not important in