Image Processing Algorithm for Fruit Identification

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 03 | Mar -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 2741
IMAGE PROCESSING ALGORITHM FOR FRUIT IDENTIFICATION
Pradeepkumar Choudhary1, Rahul Khandekar2, Aakash Borkar3 , Punit Chotaliya4
1,2,3,4 Student, Dept. of Electronics and Telecommunication Engineering,
K. J. Somaiya Institute of Engineering and Information Technology, Maharashtra, Mumbai.
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Fruits should be quickly and correctly
differentiated from their surroundings for the fruit
harvesting robot. Edge based and color based detection
methods are generally used to segment images of fruits
obtained under natural lighting conditions. In this work,
Digitized images of mango fruits along with its background
were selected from the Internet in order to find a mango in
each image and to locate its exact position. We compared
the results of Edge based and colored based segmentation
results and found that color based segmentation
outperforms the edge based segmentation in all aspects. The
comparison results are shown in the segmented image
results. Accordingly, a new mango detection method is
proposed to position the centroid of mangoes.
KeyWords: color based segmentation, edge based
segmentation, machine vision, clustering, OpenCV.
I. INTRODUCTION
In the last decades, researchers have interested in fruit
detection algorithms and applied many different computer
vision techniques. The overall aim of most of these studies
is robotic fruit harvesting [1]. But fresh fruit harvesting is
a sensitive operation. According to [2], cost of harvesting
by labors is very expensive and time consuming. In
addition, picking of fruits by hand is very tedious. To solve
these problems, human works can be replaced by
automatic robots. Automatic harvesting operations reduce
the harvesting costs [3]. For automatic fruit/vegetable
harvesting systems, it is extremely important to effectively
detect the object in outdoor conditions. There are
several problems on fruit detection in outdoor condition,
which can be classified into two groups: lighting and
occlusion. Overcoming these problems is very crucial for
the success of robotic harvesting. The first major task of a
harvesting robot is to recognize and localize the fruit on
the tree. This paper focuses on recognition of mango fruits
by using edge and color based segmentation methods and
we compare the results of both segmentation results. In
the next section the details of our proposed edge and color
based segmentation methods are presented. The results
and discussion is given in section III. Finally, in section IV,
conclusions of the proposed approach were presented.
II. DETECTION OF MANGO FRUIT ON
TREE:
A. Color based segmentation
Proposed Model:
Pre-processing input images:
In this step, we perform some necessary operators on the
captured image. The captured image of size m*n is
converted into square image of size 256*256 fig (a).
Image segmentation using K-means
clustering Algorithm:
The image segmentation means that the original image is
divided into desired number of parts. Its purpose is to
cluster the pixels which have the same features in image. It
is the first step of process. There are several methods for
this aim. Some of the works are based on the histogram or
clustering. In detection of fruit, it is purposed to separate

the fruits from background in the image. Here, the method
which has been preferred to use is 'k means clustering
algorithm’, K means clustering algorithm is superior from
the other clustering algorithm and is known as
unsupervised clustering algorithm [1]. With K-mean,we
obtained k different groups from data sets. The distances
between every groups are calculated. So, that the groups
have to be far away as much as possible from the groups
which are not similar, and the members of the particular
group are somewhat similar.
K means clustering algorithm requires iteration. For this
method, firstly, clustering centres are determined and the
data is clustered according to those centers. New
clustering centers are assigned to previous results. As
repetition of these processes is increased,result of the
algorithm is close to perfection. In the image processing, k
means clustering algorithm provides less number of color
sets for current colors of image fig(b).
Image Binarisation:
Segmented image is converted into binary image fig(c).
Image binarization is the process of converting rgb image
into the binary image. A binary image is a digital image
that has only two possible values for each pixel. Binary
images often arise in digital image processing as masks or
as the result of certain operations such as segmentation,
thresholding, and dithering x.
Morphological operations:
The Binary image is morphologically opened fig(d) using
opening function from OpenCV. A disc- shaped structural
element is used. Using this disc shaped structural element,
removal of small blobs takes place. Here the small regions
which do not belong to fruit can be eliminated by using
threshold value. If the number of pixels in this region less
than the threshold value, they can be deleted fig(e), thus
the unwanted region turns black after this operation. The
erosion is performed after removal of small objects. The
erosion makes the edges sharp fig(f).
Finding centroid of object:
After performing morphological operations, the center is
calculated as follows:
Xdistance = Xmax – Xmin
Ydistance = Ymax – Ymin
center point = [(Xmin + Xdistance/2), (Ymin +
Ydistance/2)].
After these values are calculated the geometric center of
estimated mango fruit was positioned as shown in fig (g).
(a) Original image (b) Segmented image
(c) Binary image (d) Opening operation
(e)Removal of small (f) Erosion operation
objects
(g)Mango position with center positions
Fig. 1. IMAGE SEGMENTATION ALGORITHM
IMPLEMENTATION
B. Edge Detection
The Canny edge detector is a popular method for finding
edges that begins by smoothing an image by convolving it
with a Gaussian of a given sigma value. Based on the
smoothed image, derivatives in both the x and y direction
are computed; these in turn are used to compute the
gradient magnitude of the image. Once the gradient
magnitude of the image has been computed, a process

called 'non-maximum suppression' is performed; in which
pixels are suppressed if they do not constitute a local
maximum.
The final step in the canny edge detector is the
hysteresis operator, in which pixels are marked as either
edge, non-edges and in-between, this is done based on
threshold values. The next step is to consider each of the
pixels that are in-between, if they are connected to edge
pixels these are marked as edge pixels as well. The result
of this edge detector is a binary image in which the white
pixels closely approximate the true edges of the original
image. The input image (fig 1.a) and edge detected images
is shown in fig 2.
Fig 2. Canny Edge Detector Result
III. RESULTS AND DISCUSSION
The main idea was to develop a general algorithm under
various natural lighting conditions. Thereby, no
supplemental lighting source was used to control the
luminance. Since the images were acquired under
uncontrolled natural daylight conditions, they included
tree canopies including tree branches, leaves, fruits, sky,
etc. Each object of the image has its own edges, making
image sets of edges of which the mango is just a subset. So,
an edge detection algorithm was not successful (Fig. 1).
Color based algorithm detect the fruit regions in the
images better; however, it was more complicated than the
edge detection. It can be safely concluded that this method
can achieve reliable and accurate results. The accuracy of
the color based algorithm was 85%.Figure 7 shows the
comparative analysis of proposed edge and color based
segmentation.
IV. CONCLUSION
In this paper, the segmentation method based on color
segmentation accurately detects the fruit regions in the
image. It outperforms the results of edge based
segmentation. So, edge detection based method was not as
reliable as color based segmentation; color based
algorithm was able to detect mangoes with 85-90%
accuracy.
In color based algorithm, Clustering method has
been applied on the images because of sharp differences
between background. But it couldn’t be achieved for all the
images. There are some reasons which have had influence
on analysis. Especially illumination conditions cause this
situation. Also, there can be observed false detection. Not
only the fruits but also background objects seem to be
different color tones because of sunlight. Therefore, it can
be detected as a fruit, even if there isn’t any fruit.
Occlusion is another problem which prevent sufficient
analysis. In image processing detection fruit is a difficult
problem. In this work, better results were obtained than
other works in the literature.
REFERENCES
[1] Bengi OZTURK, Murvet KIRCI, Ece Olcay GUNES, "On
Detection of Green and Orange Color Fruits in Outdoor
Conditions for Robotic Applications".
[2] Y. Sarig, "Mechanized fruit harvesting-Site Specific Solutions,"
Information and Technology for Sustainable Fruit and Vegetable
Production, FRUTIC vol. OS, pp. 237-247,2005.
[3] R. Thendral, A. Suhasini, and N. Senthil, "A Comparative
Analysis of Edge and Color Based Segmentation for Orange Fruit
Recognition" International Conference on Communication and
Signal Processing,2014.

Image Processing Algorithm for Fruit Identification

More Related Content

What's hot (19)

Similar to Image Processing Algorithm for Fruit Identification (20)

More from IRJET Journal (20)

Recently uploaded (20)

Image Processing Algorithm for Fruit Identification