An Expert System for Improving Sieve Calibration Process

Full text

International Journal of Computer Applications (0975 – 8887)
Volume 79 – No8, October 2013
18
An Expert System for Improving Sieve Calibration
Process
Peterson Adriano Belan
Departament of Informatics,
Universidade Nove de Julho,
São Paulo, Brazil
André Felipe H. Librantz
Industrial Eng. Post Graduation
Program, Universidade Nove
de Julho, São Paulo, Brazil
Sidnei Alves de Araújo
Industrial Eng. Post Graduation
Program, Universidade Nove
de Julho, São Paulo, Brazil
ABSTRACT
The reliability of the results obtained from instruments
calibration is a problem frequently found in the calibration
laboratories, especially when these instruments are
mechanical and do not have a built-in communication
interface. In this case, the time consuming is increased
significantly and the calibration may be subject to human
error. In this paper, a machine vision based system for
automatic calibration of sieve was presented. The proposed
equipment joined to the proposed technique showed in the
results a reduction of 97% in the spending time for calibration
process when compared to the traditional methods with the
same accuracy.
Keywords
Calibration, Computer Vision, Sieve, Otsu algorithm.
1. INTRODUCTION
A common problem found in the calibration laboratories
nowadays is the productivity of their technicians, the
reliability of the results obtained in the calibration process and
the transcript of the calibration certificate [1]. This is mainly
caused by the fact that, the technician who carried out the
calibration service does not issue the certificate, but only signs
it after the filling done by a typist. In some cases, the result in
the certificate can be different that one obtained during the
calibration process, frequently caused by human errors in
transcription data.
The calibration processes accuracy may be significantly
affected when instruments without built-in communication
interface are involved because the several manuals readings
and transcriptions of the data are more subject to human
errors. Moreover, this task is time consuming and stressful [1,
2]. There are two common ways to calibrate sieves: physically
examine each sieve using a profile projector and conduct
calibration tests using glass sphere, however the second way
doesn’thavethesameprecisionasthefirstone,becauseinthe
first it’s possible to measure the sieve mesh and the wire
thickness, while the second way only provides a qualitative
results.
Therefore, computer vision systems play a very important role
nowadays in the calibration of measuring instruments because
they provide greater accuracy, repeatability and cost savings,
beyond the reduction of monotonous and complex tasks [1-3].
Computer vision can be defined as a subarea of image
processing that studies the development of methods and
techniques that enable a computer system to recognize objects
in images imitating some capabilities of the human visual
system, as the ability to describe a scene contained in a digital
image.
Indeed, an efficient computer vision system must be able to
extract a set of attributes that accurately describes a scene and
small enough to reduce the spending processing time to be
applied used in practical applications such as robot vision
systems, autonomous vehicles, surveillance systems,
automatic license plate recognition, industrial inspection and
biometrics patterns recognition.
In the last years, many authors have proposed automatic
calibration approaches using computer vision techniques [1-
9]. Despite of the large number of applications, few results
have been already reported in the literature about automatic
sieve calibration.
In this context, a machine vision approach is presented in this
work, to automate the process of calibration of sieves making
the measure of the sieve mesh and its wire thickness. For that,
an algorithm using Otsu binarization [10] and the connect
component [10] method were proposed. The calibration
values could be stored automatically in a database, reducing
the possibility of human errors in data reading/ transcription
process for the issuance certificate. The remainder of the
paper is organized, as follows. Section 2 describes the
equipment developed for the experiments. In section 3, the
proposed method for the calibration of the sieve is described.
In the section 4, some experimental results are shown and
finally, the Section 6 concludes the paper, relating further
investigations.
2. MATERIALS AND METHODS
2.1 Developed equipment for sieve
calibration
The device for calibration of the sieve provides rotation for
the sieve, so that this displacement could run the complete
revolution in the sieve during the calibration process, stopping
five times on its course to acquire images of the mesh that
allows sampling compatible with that one required by the
rules quality of calibration laboratories.
This device (Figure 1) requires a special lighting. It uses a
light below the sieve and a second light above the sieve,
provided by the illumination of the USB microscope. The
device comprises equipment a standard camera supports,
lighting table, a USB microscope with 500x magnification
with LED illumination, a rotating shaft, 1W LED light
polarized, a stepper motor and a driver for controlling it.

International Journal of Computer Applications (0975 – 8887)
Volume 79 – No8, October 2013
19
Fig 1: Equipment developed for the automatic sieve calibration
2.2 Image acquisition
In this work, twenty images were acquired of two mesh
sieves: a new one and a used sieve. As can be seen in Figure
2a the first one presents mesh clean, without dirt and
completely uniform. On the other hand, the used sieve
presents in its mesh considerable changes in hole size and
problems of dirt, as well, normally caused by continuous use,
as shown in Figure 2b.
(a)
(b)
Fig 2: Example of acquired images of the sieves (a) a
new sieve without use and (b) used sieve
As reference, it was used an acquired image of a rule
graduated crystal pattern. This image was used to calibrate the
system by measuring the number of pixels between the lines
shown in Figure 3.
Fig 3: Graduated scale pattern.
3. SIEVE AUTOMATIC CALIBRATION
SYSTEM
The calibration consisted in measuring the spaces on the
sample mesh. This reading must take into account five
different regions of the sieve and analyze at least 600 holes in
the mesh. In this work, a sieve with the mesh of 38μm has
been used. The Figure 4 illustrates the calibration of the sieve-
process.
The proposed system is composed by following steps:
 Image acquisition: in this step an image of the mesh is
acquired using the USB microscopic with magnification
varying from to 5x to 500x;
 Pre-processing: Otsu algorithm threshold and a filter to
discard 4-connectedcomponents were applied to reduce
noise, i.e. dirt that could interfere in the results;
 Measurement: a connected components analysis was
carried out to measure the size of the hole and a

International Journal of Computer Applications (0975 – 8887)
Volume 79 – No8, October 2013
20
projection histogram was applied to determine the wire
thickness;
 Save the data: this step consist in saving in a database the
read data from measurement.
The result of the preprocessing is shown in Figure 5. Figure
5a displays the acquired image, Figure 5b represents the
application of the Otsu and finally Figure 5c shows the last
step of the preprocessing.
Fig 4: Steps of the proposed method
(a)
(b)
(c)
Fig 5: Preprocessing process. (a) original image was acquired using zoom of 500x, (b) Result of Otsu algorithm applied to the
Figure 5a and (c) result after applying the filter of connect components to Figure 5b.

International Journal of Computer Applications (0975 – 8887)
Volume 79 – No8, October 2013
21
One can observe that the after the pre-processing step the
remaining points to be analyzed are ten times higher than the
recommended in the manual calibration process.
After that, the analysis of the hole sizes is conducted. It
consists in of 8-connected-components counting, measuring
the area of each component (hole sieve) and comparing this
value with that one obtained by calibration of the system
based on a standard scale (Figure 3).
Subsequently, the thickness of wire is determined, rotating the
image from -45 degrees to 45 degrees, with 0.5 degree step.
For each rotating image, vertical and horizontal projection
histograms are calculated, and the highest value is assigned as
a point for measurement. The result of this process is shown
in Figure 6.
Fig 6: Result of the thickness of wire measured by the proposed method
All values acquired during the calibration process, are stored
in a database to issue the calibration certificate. The software
has been developed in C++ language using the image
processing library PROEIKON [11].
4. RESULTS AND DISCUSSION
Six hundred holes per region were analyzed, i.e. 3000 holes
on average per sieve, which is ten times larger than the sample
of the traditional method, currently carried out using profile
projector equipment. Moreover, the traditional method spends
at least 3 hours while the proposed system can reduce this
process time at ~36 times (5 min). The spent time and the
accuracy of the proposed system have been compared with
values acquired in a laboratory subject to national rules of
calibration. Figure 7 shows the distribution of obtained results
in the calibration of one sieve. One can see that they are in
good agreement with those obtained by traditional method,
thus ensuring the reliability of the system.
The Figure 8 shows the interface of the automatic sieve
calibration system, in which is possible to view the program
resume output.

International Journal of Computer Applications (0975 – 8887)
Volume 79 – No8, October 2013
22
Fig 7: Distribution of the obtained results by the proposed system
Fig 8: Interface of the sieve calibration system
The results showed in average, deviations of 1.1 µm, which is
considered acceptable, validating the results and applicability
of the system in calibration laboratories.
5. CONCLUSIONS
In this study a machine vision based system for automatic
sieve calibration was developed. The results indicate that the
system was effective for measurement of the holes and the
wire thickness. The steps of preprocessing and recognition are
extremely fast, which makes the technique robust and
efficient for calibration tasks, as it allows a significance
improvement in calibration process and the certificate
issuance, reducing the spending time in 97% compared to the
manual technical of measurement. The results indicate that the
developed system can be an interesting alternative to be used
in calibration laboratories.

International Journal of Computer Applications (0975 – 8887)
Volume 79 – No8, October 2013
23
6. ACKNOWLEDGMENTS
The authors would like to thank Uninove by financial support
and CAPES for the scholarship granted to one of the authors.
7. REFERENCES
[1] F.C. Alegria,A.C. Serra, Computer vision applied to the
automatic calibration ofmeasuring instruments,
Measurement.28 (2000)185–195.
[2] F.C. Alegria, A.C. Serra,Automatic Calibration of
Analog and Digital Measuring Instruments Using
Computer Vision, IEEE Transactions on Instrumentation
and Measurement.49 (2000) 94–99.
[3] G. Andria et al., Automatic Calibration System for
Digital Instruments Without Built-In Communication
Interface,Anais do XIX IMEKO World Congress
(Lisbon). (2009) 857–860.
[4] E.V. Fernández et al., A machine vision system for the
calibration of digital thermometers, Measurement
Science and Technology. 20 (2009) 1–7.
[5] B. Hemming, H. Lehto, Calibration of dial indicators
using machine vision, Measurement Science and
Technology. 13 (2002) 45–49.
[6] B. Hemming, A. Fagerlund, A. Lassila, High-accuracy
automatic machine vision based calibration of
micrometers, Measurement Science and Technology. 18
(2007) 1655–1660.
[7] W.He et al., Intelligent reading method for analog meter
based on computer vision, Chinese Journal of Scientific
Instrument. 28 (2007) 1234–1239.
[8] P.A. Belan, S.A. Araújo, A.F.H. Librantz, Computer
vision techniques applied to the process of calibrating
measurement instruments having numeric displays
without built-in data communication interfaces, Exacta.
10 (2012) 83–91.
[9] P.A. Belan, S.A. Araújo, A.F.H. Librantz, Segmentation-
free approaches of computer vision for automatic
calibration of digital and analog instruments,
Measurement Science and Technology. 46 (2013) 177–
184.
[10] R.C. Gonzalez, R.E. Woods, Digital Image
Processing.Massachusetts: Addison-Wesley. 793 pp,
2002.
[11] H.Y. Kim, ProEikon –Library for Image Processing and
Computer Vision <http://www.lps.usp.br/~hae/software>
Access at fev. 2010.
IJCATM : www.ijcaonline.org