The working process of the automated strawberry grading system
is described as the following:
(1) Strawberry is put on the conveyer belt by manual, advanced
with the conveyer belt at a constant speed.
(2) When the first photoelectrical sensor detects the strawberry
to be passing through, it sends the signal to the control part
and begins to capture the strawberry images, extract the grading features and determine which gradation the strawberry belongs to.
(3) As the strawberry’s movement continues, when the second
photoelectrical sensor detects the strawberry to pass through,
the control part controls the mechanical parts to implement
corresponding grade according to the result of the image processing.
(4) Repeat the step (1), until all strawberries were detected.
2.2. Strawberry feature extraction and gradation algorithm
(1) The strawberry position determination
The strawberries are placed on the conveyer belt at random,
so the strawberry’s positions are uncertain in the images captured
by computer. Because there are some calyxes at the stem,
the fruit position can be determined after these calyxes are
found.
In Fig. 2(d), the g–r value is different among the strawberry
calyx, fruit and the background. The g–r gray image converted
from the colour image is shown in Fig. 2(b), it shows that the g–r
value of the calyx is greater than that of the fruit and the background,
so the image can be segmented by selecting a threshold.
And the extracted strawberry calyx is shown in Fig. 2(c).
(2) Strawberry gradation by the single feature
The fruit gradation generally uses size, shape and colour.
The automated strawberry grading system mainly extracts the
three features, and implements corresponding gradation.
(1) Strawberry shape feature
The strawberry shape is too complicated to measure with
a single geometry size. This paper found that the strawberry
shape feature parameters can be obtained by extracting line
sequences from the strawberry contour and normalizing
the length of these line sequences to eliminate the influence
of the strawberry size, and these parameters could express
the strawberry shape well. In order to make the shape
gradation adapt various kinds of strawberry, and have a
faster processing speed to meet the real-time requirement,
the automated strawberry grading system implemented Kmeans
clustering method to complete the shape gradation.
K-means clustering method, put forward by Marques et
al. (2002), is the optimal algorithm in clustering analysis.
According to the advance class centers, theK-means clustering
method could put some similar classes into one center.
It is described as follows: classify n objectives into K classes
with the parameter K, having a higher similarity within one
class and a lower similarity among classes. This method has
certain intelligence, so it can select automatically the center
of a class and the measurement threshold of the similarity
according to the sample’s variety. Therefore, it can be
applicable to various strawberry gradations.
The steps to extract the strawberry shape features are
described as the following:
(a) Select the R–G channel of the strawberry image
(Fig. 3(a)); select a segmentation threshold based on
Outs algorithm. This method can avoid a great deal
mathematic calculation caused by the colour space
transform, and achieve better segmentation result
(Fig. 3(b)). The segmented binary image is traced for
the contour and then the strawberry’s contour curve
(Fig. 3(c)) is obtained.
(b) Make the fruit contour be added with the calyx contour
to obtain the joining curve between the calyx and
the fruit. The joining curve center connected with the
fruit gravity center to form a line l1, i.e., the major axis
direction of the strawberry.
The working process of the automated strawberry grading systemis described as the following:(1) Strawberry is put on the conveyer belt by manual, advancedwith the conveyer belt at a constant speed.(2) When the first photoelectrical sensor detects the strawberryto be passing through, it sends the signal to the control partand begins to capture the strawberry images, extract the grading features and determine which gradation the strawberry belongs to.(3) As the strawberry’s movement continues, when the secondphotoelectrical sensor detects the strawberry to pass through,the control part controls the mechanical parts to implementcorresponding grade according to the result of the image processing.(4) Repeat the step (1), until all strawberries were detected.2.2. Strawberry feature extraction and gradation algorithm(1) The strawberry position determinationThe strawberries are placed on the conveyer belt at random,so the strawberry’s positions are uncertain in the images capturedby computer. Because there are some calyxes at the stem,the fruit position can be determined after these calyxes arefound.In Fig. 2(d), the g–r value is different among the strawberrycalyx, fruit and the background. The g–r gray image convertedfrom the colour image is shown in Fig. 2(b), it shows that the g–rvalue of the calyx is greater than that of the fruit and the background,so the image can be segmented by selecting a threshold.And the extracted strawberry calyx is shown in Fig. 2(c).(2) Strawberry gradation by the single featureThe fruit gradation generally uses size, shape and colour.The automated strawberry grading system mainly extracts thethree features, and implements corresponding gradation.(1) Strawberry shape featureThe strawberry shape is too complicated to measure witha single geometry size. This paper found that the strawberryshape feature parameters can be obtained by extracting linesequences from the strawberry contour and normalizingthe length of these line sequences to eliminate the influenceof the strawberry size, and these parameters could expressthe strawberry shape well. In order to make the shapegradation adapt various kinds of strawberry, and have afaster processing speed to meet the real-time requirement,the automated strawberry grading system implemented Kmeansclustering method to complete the shape gradation.K-means clustering method, put forward by Marques etal. (2002), is the optimal algorithm in clustering analysis.According to the advance class centers, theK-means clusteringmethod could put some similar classes into one center.It is described as follows: classify n objectives into K classeswith the parameter K, having a higher similarity within oneclass and a lower similarity among classes. This method hascertain intelligence, so it can select automatically the centerof a class and the measurement threshold of the similarityaccording to the sample’s variety. Therefore, it can beapplicable to various strawberry gradations.The steps to extract the strawberry shape features aredescribed as the following:(a) Select the R–G channel of the strawberry image(Fig. 3(a)); select a segmentation threshold based onOuts algorithm. This method can avoid a great dealmathematic calculation caused by the colour spacetransform, and achieve better segmentation result(Fig. 3(b)). The segmented binary image is traced forthe contour and then the strawberry’s contour curve(Fig. 3(c)) is obtained.(b) Make the fruit contour be added with the calyx contourto obtain the joining curve between the calyx andthe fruit. The joining curve center connected with thefruit gravity center to form a line l1, i.e., the major axisdirection of the strawberry.
การแปล กรุณารอสักครู่..