Sei sulla pagina 1di 7

IDL - International Digital Library Of

Technology & Research


Volume 1, Issue 4,April 2017 Available at: www.dbpublications.org

Internati onal e-J ournal For Technol ogy And Research-2017

Ripeness Evaluation of Mango using Image


Processing
Sharadha shree 1 , Dr. Chandrappa D.N 2
1
Student ,M.Tech, DEC, GMIT Davangere.
2
Professor and HOD, Dept. Electronics Engineering, GM IT DA VANGERE.

ABSTRACT 1. INTRODUCTION
Abstract Mangoes are delicious seasonal fruits Agriculture is one of the largest economic sector and it
grown in the tropics. They are harvested from i ts plays a major role in the development of our country. Th e
grove when matured enough for the market. They do ever-increasing population demands for higher quality of
not mature uniforml y in trees but stage by stage. mangoes with good appearance. There is a need for the
Most farmers use manual experts for ri pening development of accurate, fast and focused quality
evaluati on of the mang oes which is ti me consuming, determination of mangoes.
inconsistent and inaccurate . To avoi d manual effort, Harvesting of mangoes is performed in several steps like
an automated Computer vision techni que is cutting of fruits from the farm, washing, sorting, grading,
introduced in this paper. This includes preprocessing, packing, transporting and finally storing. Out of these
Segmentation, Feature extraction and classification. sorting and grading are major processing tasks associated
Here 24 color features are extracted from the mang o for preserving the quality of mangoes.
image. Classifying the mango into two di fferent Sorting of mangoes is done based on appearance of fru its,
classes according to their maturity level using k-NN whereas grading is done based on the overall quality
Classifier and this proposed system resulted i n the features of fruits by considering a number of attributes
accuracy of about 97% in evaluati on of ri peness of like shape, size, color and texture etc.

Mangoes. Classification is necessary for the quality evaluation of


mangoes. Fruit industry for its excellent trading purpose
goes for highly selective ones in quality and standard. It
General Terms
demands the suppliers and the distributors the fruits of
Quality Inspection, Feature Extraction, Color Feature.
high standards of quality, package and presentation. So

Keywords there is a increasing need to supply quality fruits within a


short period of time has development of Co mputer Vis ion
Feature Extraction KeywordsSegmentati on, Techniques to improve the quality.

Feature Extraction, k-NN classifier. In present scenario, sorting and grading of fruit
according to maturity level are performed manually
before transportation. This manual sorting by visual
inspection is labour intensive, time consuming and

IDL - International Digital Library 1 |P a g e Copyright@IDL-2017


IDL - International Digital Library Of
Technology & Research
Volume 1, Issue 4,April 2017 Available at: www.dbpublications.org

Internati onal e-J ournal For Technol ogy And Research-2017

suffers from the problem of inconsistency and Tajul Rosli et al. Proposed and implemented
inaccuracy in judgement by different hu man and methodologies and algorithms to determine the grade of
provides an opportunity to apply Computer Vision local mango production in Perlis. The main contribution
based system to assess this problem a design and development of an efficient algorith m for
detecting and sorting the mango at more than 80%
accuracy in grading co mpared to hu man expert sorting.
Problem statement This work proposes a mango grading technique for

For exports, grading is done manually by the experts now mangoes quality classification by fuzzy logic based

which is very time consuming and subjective. So, image processing [3].

farmers need alternatives for sorting and grading P. Sudhakara Rao et al. (2009) have adopted HSI

mangoes. An automated mango sorting system could be model for sorting and grading of fruits by color and

more preferable as it can be cheaper, consistent and developed a system for on-line sorting of Apples based

could result in better overall quality. Thus the scope of on color, size and shape. Images are captured by a color

the project is to develop an automated computer vision CCD camera and frames are separated by a frame grabber

system for sorting of mango based on maturity level. card and it produced the image in RGB model. The image
is analyzed by using advanced image processing

Literature Survey techniques to estimate the color of image. Using image


processing system achieved around 98 % accuracy in
Chandra Nandi et al. (2012), imp lemented a computer
color inspection of apples [4].
vision based system for automatic grading and sorting of
Suzanawati Abu et al. Proposed and imp lemented
mangoes based on maturity level fro m its RGB image
Automated Mango Fruit Assessment Using Fuzzy Logic
frame, collected with the help of CCD camera.
Approach. This work developed a new method of
Parameters of different classes of mangoes are estimated
automated mango Size and grade assessment using RGB
using Gaussian Mixture Model. Graph contour tracking
fiber optic sensor and fuzzy log ic approach. The
method based on chain code is adapted for finding the
calculation of maximu m, min imu m and mean values
boundary of the mango. This automated technique is
based on RGB fiber optic sensor. To analyse the data and
good but is further affected by ambient light intensity.
make the classification for the mango fruit uses the
Response time of system is on the order of 50 ms [1].
minimu m entropy formu lation method. The automated
CCD camera was used to collect video image of
mango grading system using fuzzy logic achieved
Mangoes and several significant features of maturity
77.78% accuracy in overall categories [5].
level of Mango was obtained. Colour of Mango was
estimated using Gaussian Mixture Model. Accuracy can
Methodology
be improved by using Support vector machine and neural
network. Gaussian Mixture Model (GMM ) and fuzzy The follo wing steps are followed to perform the ripeness

logic was comb ined for size based grading of Mango. evaluation of mango

Size of Mango was calculated using pixel area covered 1. Select the mango image.

by Mango [2]. 2. Crop the area of mango.


3. Find and calcu late the major axis.

IDL - International Digital Library 2 |P a g e Copyright@IDL-2017


IDL - International Digital Library Of
Technology & Research
Volume 1, Issue 4,April 2017 Available at: www.dbpublications.org

Internati onal e-J ournal For Technol ogy And Research-2017

4. A lign the mango image. Fig 1 .Fl owchat


5. Divide the 3 regions.
6. Extract the color features fro m the 3 reg ions.
7. Use nearest neighbour classification algorith m for
database features.
8. Output results based on the maturity level.

1) The basic global threshold, T, is calculated as


follows:

1. Select an init ial estimate for T


(typically the average grey level in the
image)

2. Segment the image using T to produce


two groups of pixels: G1 consisting of
pixels with grey levels >T and G2
consisting pixels with grey levels T

3. Co mpute the average grey levels of


pixels in G1 to give 1 and G2 to give
2

4. Co mpute a new threshold value:

5. Repeat steps 2 4 until the difference


in T in successive iterations is less than
a predefined limit T.

IDL - International Digital Library 3 |P a g e Copyright@IDL-2017


IDL - International Digital Library Of
Technology & Research
Volume 1, Issue 4,April 2017 Available at: www.dbpublications.org

Internati onal e-J ournal For Technol ogy And Research-2017

Fig 2. Input Mango i mage Fig 3. Conversion to g rays ca le

Fig 4. Image after threshol ding Fig 5. Centroi d of mang o (red * )

IDL - International Digital Library 4 |P a g e Copyright@IDL-2017


IDL - International Digital Library Of
Technology & Research
Volume 1, Issue 4,April 2017 Available at: www.dbpublications.org

Internati onal e-J ournal For Technol ogy And Research-2017

After the analysis it was found that categorizat ion into


unripe, semi ripe and ripe is possible by the range as in So at this point the dominant color, spots and size features are
extracted from the mango. All the features with corresponding
Table 1. values are shown in Fig 6. These are the feature parameters for
the grading. Based on this the grading rules are created. Some
of them are shown Table 2.
Table 2: S ample rules of grading

Rule 1 Rule 4
If spots are more and If spots are average and
Size is less and Size is medium and
Color is unripe then Color is unripe then
Fig 6. Dominant bin in histogram Grade is Rejected Grade is Grade 3
The following range of the dominant bin over a channel of
Lab color space was observed
Table 1: Dominant a color range Rule 8 Rule 10
If spots are less and If spots are less and
Dominant a Category Size is medium and Size is large
Color is semi ripe then Color is Ripe then
a<=117 Unripe Mango Grade is Grade 2 Grade is Grade 1
117<a<130 Semi Ripe Mango

a>=130 Ripe Mango


On the basis of grading rules results are shown in section 3.

IDL - International Digital Library 5 |P a g e Copyright@IDL-2017


IDL - International Digital Library Of
Technology & Research
Volume 1, Issue 4,April 2017 Available at: www.dbpublications.org

Internati onal e-J ournal For Technol ogy And Research-2017

Table 3. Accuracy with the proposed method as well as efficient. In the future work the texture on
the surface of the mango can also be used as a feature
Correctly Total Accuracy
parameter fo r grading so that the overall accuracy can
Classified Correct
Grade Proposed be improved.
S amples S amples
Extra Class 86 86 Method
100%
Acknowledgemet
Grade 1 55 58 94.82%
Authors would like to thank Mr.Ruknuddin Kazi
Grade 2 187 196 95.40%
(Valsad, Gu jarat) fo r the authentication of the dataset of
Rejected 52 58 89.66%
the mangoes.
Average Accuracy 94.97%

5. REFERENCES
Table 4. Comparison of time [1] Xu Liming and Zhao Yanchao, "Automated
strawberry grading system based on image
Full Mango Proposed Method
processing," Computers and Electronics in
Time (seconds) 499.43 secs 377.23 secs
Agriculture, vol. 71, no. Supplement
Average time for
1.248 sec 0.9425 sec 1, pp. S32-S39, April 2010.
1 mango

[2] Taju l Rosli Bin Razak, Mahmod Bin


4. CONCLUSION AND FUTURE Othman(DR), Mohd Nazari Bin Abu Bakar(DR),
DIRECTION Khairu l Adilah BT Ahmad, and AB.Razak Bin
An algorithm for the grading of the mangoes was Mansor, "Mango Grading By Using Fu zzy Image
developed. Color, spots and size were taken as the Analysis," in In proceedings of International
feature parameters for the grading the mangoes into Conference on Agricultural, Environ ment and
grade1, grade2, grade3 and Bio logical Sciences, Phuket, 2012.
rejected. A modified dominant color feature
ext raction [3] Teoh, Yeong Kin, Abu Hasan, Suzanawati, AND
technique using the Lab color model had been proposed. Sauddin@Saduddin, Suraiya. "Automated Mango
Instead of taking the whole mango the color feature was Fruit Grading System Using Fu zzy Logic"
ext racted from the strip on the surface of mango. The Journal of Agricultural Science [Online],
spots were extracted fro m the whole surface of the Volu me 6 Nu mber 1, December 2013.
mango. Length of the major axis was used for the size
feature. Experiment shows that the proposed method [4] Kazuhiro Nakano, "Application of neural networks
has more accuracy as well as efficiency. to the color grading of apples," Computers and
Electronics in Agriculture, vol. 18, no. 2-3, pp. 105-
Proposed algorithm was able to classify the mangoes 116, August 1997.
with an accuracy of 94.97 % into various grades. And
time taken for the grading of a mango on average was
less than a second. So the proposed method is accurate
IDL - International Digital Library Of
Technology & Research
Volume 1, Issue 4,April 2017 Available at: www.dbpublications.org

Internati onal e-J ournal For Technol ogy And Research-2017

[5] M. Khojastehnazhand, M. Omid, and A. evaluation of Iyokan orange fruit using neural
Tabatabaeefar, "Development of a lemon sorting networks," Computers and Electronics in
system based on color and size," African Journal of Agriculture, vol. 29, no. 1-
Plant Science, vol. 4(4), pp. 2, pp. 135-147, October
122-127, April 2000.
2010.
[11] Ankur M Vyas, Bjial Talat i and Sapan Naik.
[6] Akira Mizushima and Renfu Lu, "An image Article: Co lour Feature Extraction Techniques of
segmentation method for apple sorting and grading Fruits: A Survey. International Journal of
using support vector machine and Otsus method," Computer Applications 83(15):15-22, December
Computers and Electronics in Agriculture, vol. 94, 2013. Published by Foundation of Co mputer
pp. 29-37, June Science, New York, USA.
2013.
[12] Cheng, Heng-Da, X. H. Jiang, Ying Sun, and
[7] Nursuriat i Jamil, Azlinah Mohamed, and Syazwani Jingli Wang. "Colour image segmentation: advances
Abdullah, "Automated Grading of Palm Oil Fresh and prospects." Pattern recognition 34, no. 12
Fruit Bunches (FFB) using Neuro -Fuzzy (2001): 2259-
Technique," 2009 2281
International Conference of Soft Computing and .
Pattern
Recognition, pp. 245-249, [13] Naik S. and Patel B, CIELab based color
2009. feature extraction for maturity level grading
of Mango (Mangifera Indica L ), National journal
[8] Yousef Al Ohali, "Co mputer vision based date of System and Information technology , ISSN: 0974
fruit grad ing system: Design and implementation," 3308.
Journal of King Saud University - Computer and
Information Sciences, vol. 23, no. 1, pp. 29-39,
January 2011

[9] U. Ahmad, M . Suhil, R. Tjah johutomo, and


H.K.
Purwadaria, "Develop ment of Citrus Grading System
Using Image
Processing". [Online]
http://repository.ipb.ac.id/handle/123456789/68
910

[10] Naoshi Kondo, Usman Ahmad, Mitsuji Monta,


and Haruhiko Murase, "Machine vision based quality

Potrebbero piacerti anche