A Non-Contact Technique for Qualitative Analysis of Crayon Packets

Indian Journal of Science and Technology, Vol 8(24), DOI: 10.17485/ijst/2015/v8i24/84417, September 2015
ISSN (Print) : 0974-6846
ISSN (Online) : 0974-5645
A Non-Contact Technique for Qualitative Analysis of
Crayon Packets using Image Processing Techniques
K. V. Santhosh* and Bhagya R. Navada
Department of Instrumentation and Control Engineering, Manipal Institute of Technology, Manipal - 576 104,
Karnataka, India; [email protected]
Abstract
Packaging plays a vital role in protection of materials, along with providing information about the product. But many a
times it is difficult to identify the quality of product inside the packet. Proposed work discusses a technique to identify
the quality of such packed products. Objectives of the proposed work are to count number of crayon in each packet and
check if its’ as per desired number using the techniques of image processing. A camera is used to capture images of crayon
packets in the production line packaging section. On capturing the images it is processed online using LabVIEW platform.
Image processing algorithms like pattern recognition, edge detection and center of mass to compute the quantified data of
the image acquired. Neural network algorithm is employed to detect the faulty packages as also to display the position of
missing crayon. The system once designed is subjected to test and the results obtained proved successful implementation
of set objectives.
Keywords: Automation, Image Processing, Qualitative Analysis
1. Introduction
Montessori teaching has had a profound influence on the
education of young children through the world. Many
effective techniques are been developing for increasing
the knowledge of children. One such development is
extensive use of colors. Thus increases the uses of more
and more crayons, color pencils and so on. Over the
years production and marketing of these products have
been increasing exponentially. Larger volumes are being
produced with efficient use of automation, but there exists
a lot of flaws. It can be seen that we hear news of missing
one or two crayons in a packet, but it is difficult to notice
that while purchasing the product as these packet will
be closed. Industrial processes often rely on sample test
by humans and density check. Both these technique are
often less accurate and decrease the time of production.
Several more researchers have been working in the
field of quantitative fault analysis using several improved
technologies like in1, an electro-mechanical system is
proposed for classifying marble slabs automatically when
* Author for correspondence
they run on conveyor belt. Clustering method is used
for classification by acquiring digital images of marble
slabs and feature extraction of these images. A method
for estimation of coal particle on a conveyor based on
their size is proposed in2. A digital image of coal particle
is taken, features are collected based on its texture and
classification is carried out using support vector machine
models. In3, a method for tracking of moving objects
is reported for a robot using a vision system which can
improve robot performance in tracking, picking and
classifying objects on a conveyor. A method for detecting
2D profile and motion of a product moving on a conveyor
using orthogonal pair of line array scanners called lateral
optical sensors is reported in4. An integrated system for
object segmentation, tracking, feature extraction and
classification of four moving objects including cars,
human beings, bicycles and motor cycles is proposed
in5. Hierarchical support vector machine is used for
classification of objects. In6, a method for classification
of bag, human, body organs, group of people and clutter
using image feature extraction and many classification
A Non-Contact Technique for Qualitative Analysis of Crayon Packets using Image Processing Techniques
techniques is reported. Feature extraction from a tactile
image for classification and matching is proposed in7. 10
different object shapes are acquired. Euclidian distance
is used as similarity measure for matching and for
classification k-nearest neighbor, Naive Bayes classifier
and Linear Discriminant Analysis methods are used.
In8, an algorithm is developed for detection of 4 different
shaped objects placed on a circular conveyor during
the motion of conveyor. Design of a low cost machine
vision system is reported in9 for monitoring objects on
a conveyor are tested for the given specification if it is
matching then it gives a signal pass and if no then it gives
rejected signal. Detection of missing and broken tablets,
color, size, shape of individual tablets and tablet fragments
is proposed in10 using an automated visual inspection
system. In11, classification of biscuits moving on conveyor
belt based on color using SVM and Wilk’s λ analysis to
group the biscuits into four group based on the level of
baking is reported. Similarly in12, wavelet transforms are
used in identification of people from a video frame.
From the above survey it can be seen that many
algorithms have been used for detecting and counting of
the objects where as very less work is done in the area of
fault detection of crayon packets. Some of the methods used
may give satisfactory results in terms of economy but when
it comes to accuracy and simplicity some compromising is
required. In this paper an automatic method is proposed
to detect the number of crayons in a packet using image
processing and neural network algorithm on a LabVIEW
platform. Artificial Neural Network (ANN) module is
trained using a multi perceptron layer function using
artificial bee colony algorithm. Proposed method counts
the number of crayons in a packet and if the count is
not same as the desired, it shows it as defective packet,
secondly it also computes which crayon is missing, thus
making packaging easier and faster.
The paper is organised as follows: After introduction
in Section 1, Section 2 discusses the associated problem.
Section 3 deals with proposed solution. Results and
conclusions are given in Section 4. Finally, Section 5
discusses conclusions.
2. Problem Statement
Crayons after packed are as shown in Figure 1. In the
packet considered for analysis in the present work is
expected to have 25 crayon sticks of different colors.
2
Vol 8 (24) | September 2015 | www.indjst.org
But it so happens due to process error that the packets
may have 24 or 23 crayons as seen in Figure 2, a user will
have to buy defective product. This gives a bad image
about the manufacturer and thus manufacturer may lose
customers. The product if displayed like this it would not
have been a big problem but the packet is displayed in a
closed manner like as shown in Figure 3.
Figure 1. Crayon packets after packaging with accurate
number of crayons.
Figure 2. Crayon packets after packaging with less number
of crayons.
Figure 3. Crayon packets as seen by consumer.
The user will not have any idea about the product
then to believe it to be proper. Thus it is very important
to have a quantitative analysis before packing. The paper
proposes a technique using image processing algorithms,
to objective of the proposed work is given as
• Capture the image of the packet in the production
line.
• Apply image processing techniques to compare the
package with accurate one.
• If it’s found the package has less number of items than
desired.
• Find the number of missing crayons, with its position.
Indian Journal of Science and Technology
K. V. Santhosh and Bhagya R. Navada
• Initiate action of diverting the defective package from
production line.
• All the above process should be carried on in a definite time frame so as to cause no delay in production
line.
3. Problem Solution
For demonstration working of proposed quantitative
analysis conveyer system is designed as shown in Figure 4.
A video camera is placed to acquire video for processing.
Once the video is acquired it is converted to frame by
using slicing functions. LabVIEW platform is used for
the same, LabVIEW consists of two different windows
one being front panel 6 (through which human interface
is done) and block diagram 6 (used to design desired
program).
Figure 4. Experimental setup.
Front panel of the proposed technique is designed as
seen in Figure 5. Consisting of two image windows in the
first, one can visualize the actual cropped image of the
packet and in second we can see the processing image
for which density function is computed for each block
and used in neural network model. An indicator is used
display whether desired number of crayon is available
in the packet. If so, color of indicator is green else red.
A numerical indicator is used to display the number
of crayons present in the packet followed by a string
indicator displaying the position of missing crayon.
In processing, we start off by extracting two color
planes from the RGB plane (32-bit) image and converting
it to a one color plane image, say red plane (8-bit) image.
We convert it to a grayscale (8-bit) image because it
reduces the matrix size of the image and computational
time is immensely reduced as well as memory requirement
is very less.
Now, the whole sample image is not required to be
processed. We select our region of interest and mask all
the pixels that are out of our region of interest. We extract
out our masked region and we are left with only our
region of interest. Now, we select the range of pixels in
our region of interest which are bright by the operation of
cluster thresholding. After thresholding, the whole image
is highlighted with red color. The thresholding operation
further reduces the image from an 8-bit one to a binary
image. Now, we remove those defective portions which
were not highlighted in our thresholding operation using
a filter.
The filtered image is then analyzed by particle analysis
process. This process displays measurement results for
selected particle measurements performed on the filtered
image. We can use the center of mass technique of particle
analysis. Here, we analyze the center of mass for this
filtered image. If there is a defective sample, its center of
mass will be different as compared to the good sample’s
Figure 5. Front panel VI of proposed work.
Vol 8 (24) | September 2015 | www.indjst.org
Indian Journal of Science and Technology
3
A Non-Contact Technique for Qualitative Analysis of Crayon Packets using Image Processing Techniques
center of mass. In this way, we can compare the sample’s
image and find out which sample is a defective one and
for checking whether desired number of crayon is present
in the packet.
This process will help us to just identify whether the
desired number of crayons are present in the packet, but
for us along with this information we are also focused on
finding how many crayons are missing and what is the
position of the missing crayons for this we further process
the image using edge detection and Neural Network
Algorithm. Once it is found that crayon packet doesn’t
contain desired number of crayons it is subjected to edge
detection algorithm in this process edges between each
of the crayon is identified. In the next process mapping
function is used to mark the edges. Once it is done we
would have blocks equal to the number of crayons. This
will make the task of computation of position of missing
crayon simpler13,14.
Table 1. Summary of neural network model
Database
No. of neurons in
Transfer function of
MSE
R
Input
Training base
Validation base
Test base
1st layer
2nd layer
1st layer
Output layer
min
max
36x25
12x25
12x25
10
8
Softmax
Linear
3.256x10-4
0.9987
0
255
Centre of mass filter is used to compute the density
function of each block, which corresponds to the color
of the crayon. Now, Neural Network Algorithm is used to
compute the position of missing crayon. For the proposed
work neural network model trained by artificial bee colony
algorithm15,16 is used. For training the neural network
input will be the density function obtained from each of
the block, in our case 25. Target data will be the position
indicator. Once trained, output of neural network is used
as a function to switch case. Switch case is used to display
the result in the position of missing crayon. Summary of
the neural network model is shown in Table 1.
4. Result and Analysis
The proposed system was tested with different test cases
like with all crayons present, with one missing crayon,
with two missing crayon, etc. and keeping the time
constraint of production line into consideration. The
system produced satisfactory results.
Results obtained for different cases are compiled.
Results are viewed using the front panel of proposed vi, few
of them are shown in Figure 6, Figure 7, Figure 8, Figure 9
and Figure 10. Figure 6 shows result obtained for a case for
crayon packet with actual number of crayons. Similarly in
Figure 7, Figure 8, Figure 9 and Figure 10 shows the result
obtained for the crayon packets with missing crayons. It
can be seen from results shown from proposed system is
that of actual packets as seen from actual image. Output
of proposed system also displayed the actual position of
missing crayon. For testing around 130 cases were tested
with different combinations, proposed technique was able
to display information about 126 cases accurately.
Figure 6. Result obtained for correct packet.
4
Vol 8 (24) | September 2015 | www.indjst.org
Indian Journal of Science and Technology
K. V. Santhosh and Bhagya R. Navada
Figure 7. Result obtained for erroneous packet.
Figure 8. Result obtained for erroneous packet.
Figure 9. Result obtained for erroneous packet.
Vol 8 (24) | September 2015 | www.indjst.org
Indian Journal of Science and Technology
5
A Non-Contact Technique for Qualitative Analysis of Crayon Packets using Image Processing Techniques
Figure 10. Result obtained for erroneous packet.
5. Conclusion
An automated technique for checking the quality of
crayon packet is designed in this work. The objective of
the proposed work was to design a non-contact technique
for checking whether any crayon is missing in the packet
before it is packed. If any crayon is missing which is that
crayon?
The process was designed keeping in mind that the
process to be fully automatic and should not disturb
the production. Thus image processing technique was
incorporated which is a non-contact process. Further
the process was designed using LabVIEW which is one
of real time platform for design based applications. The
process was designed using image processing technique
and Neural Network Algorithm to produce results of
quantitative analysis of crayon packets but within a
definite time frame. Results produced show the effective
implementation of proposed work.
Further the proposed work can be upgraded to work
for industrial standards.
6. References
1. Alper Selver M, Akay O, Alim F, Bardakc S, Olmez M. An
automated industrial conveyor belt system using image
processing and hierarchical clustering for classifying marble slabs. Robotics and Computer-Integrated Manufacturing. 2011 Feb; 27(1):164–76.
2. Jemwa GT, Aldrich C. Estimating size fraction categories of
coal particles on conveyor belts using image texture mod-
6
Vol 8 (24) | September 2015 | www.indjst.org
eling method. Expert Systems with Applications. 2012 Jul;
39(9):7947–60.
3. Mei J, Ding Y, Zhang W, Zhang C. Fast detection, position
and classification of moving objects on production line.
Optik. 2010 Dec; 121(23):2176–8.
4. Lee KM, Foong F. Lateral optical sensor with slip detection
for locating live products on moving conveyor. IEEE Transactions on Automation Science and Engineering. 2010 Jan;
7(1):123–32.
5. Liang CW, Juang CF. Moving object classification using local shape and HOG features in wavelet-transformed space
with hierarchical SVM classifiers. Applied Soft Computing.
2015; 28:483–97.
6. Gurwicz Y, Yehezkel R, Lachover B. Multiclass object classification for real-time video surveillance systems, Pattern
Recognition Letters. 2011 Apr; 32(6):805–15.
7. Datta S, Khasnobish A, Kona A, Tibarewalab EDN, Janarthanan R. Performance analysis of object shape classification and matching from tactile images using wavelet energy
features. Proceedings International Conference on Computational Intelligence: Modeling Techniques and Applications; Kolkata, India. 2013. p. 805–12.
8. Jokesch M, Bdiwi M, Suchy J. Integration of vision/force
robot control for transporting different shaped/colored objects from moving circular conveyor. Proceedings International Symposium on Robotic and Sensors Environments;
Romania. 2014.
9. Arsalan M, Aziz A. Low-Cost machine vision system for
dimension measurement of fast moving conveyor products.
Proceedings International Conference on Open Source
Systems and Technologies. Lahore, Pakistan; 2012 Dec 2022. p. 22–7.
10. Derganc J, Likar B, Bernard R, Tomamevi D, Pernu F.
Real-time automated visual inspection of color tablets in
pharmaceutical blisters. Real-Time Imaging. 2003 Apr;
9(2):113–24.
Indian Journal of Science and Technology
K. V. Santhosh and Bhagya R. Navada
11. Nashat S, Abdullah A, Aramvith D, Abdullah MZ. Support
vector machine approach to real-time inspection of biscuits
on moving conveyor belt. Computers and Electronics in
Agriculture. 2011 Jan; 75(1):147–58.
12. Deshpande MM, Rana JG. Intelligent video surveillance
system based on wavelet transform and support vector
machine. International Journal of Computer Applications.
2012 Jun; 48(14):42–5.
13. Roy AK, Chaira T. Fuzzy image processing and applications
with Matlab. Taylor and Francsis Publishers; 2009 Nov.
Vol 8 (24) | September 2015 | www.indjst.org
14. Klinger T. Image Processing with LabVIEW and IMAQ Vision. Pearson Education, Inc; 2003.
15. Karaboga D, Gorkemli B, Ozturk C, Karaboga N. A Comprehensive survey: Artificial Bee Colony algorithm and
applications. Artificial Intelligence Review. 2014 Jun;
42(1):21–57.
16. Bonabeau E, Dorigo M, Theraulaz G. Swarm intelligence:
From natural to artificial systems. Oxford University Press;
1998.
Indian Journal of Science and Technology
7