MVA2005 IAPR Conference on Machine VIsion Applications, May 16-18, 2005 Tsukuba Science City, Japan

3-19
An Approach for Defect Detection and Classification of the Yarn Ends
for Splicing
Khal
edIssa Hi
roshiNagahashi

Int
erdi
sci
pli
nary GraduateschoolofSci enceandEngineeri ng,TokyoInst
itut
eofTechnol
ogy
R2-51,4259Nagat suta-cho,M i
dori-ku,Yokohama226-8503,Japan
Phone:+81-45-924-5478;Fax:+81-45-924-5175
E-mail:{Khaled,longb}@ isl
.ti
tech.ac.
jp

Abstract
This paper presents an automatic vision based system for twisted in the same di rection as thatofthe parentyarn.
qual ity controlof yarn ends ready for spl icing,which is [4,5]Splicing proceedsintwost ageswithtwodifferentair
aimed to establish a standard quality measure and lower blastsofdi fferentintensity.Thefi rstairblastunt wistsand
manufacturing cost. New approach for defect detection causesopeni ng ofthefreeends.Theunt wisted fibersare
and classification is presented. In this approach,features thenintermingl edandtwistedinthesamedi rectionasthat
describing the shape and surface defects are extracted ofparentyarnby anot herai rbl
ast.
and defects are classified into different classes. Examples
of defects are used to train the classification system using Presently,checking the appearance of the openi ng and
neuralnetwork. Experimentalresul ts show that a high splicing isstillaccomplished by human experts[6]which
detection and classification rate can be obtained using is time consuming and suffers from variability due to
this approach. human subjectivity. Consequently, automated
investigation ofyarn endsand automated classifi cation of
splice joint faults are highl
y desi rable.Such a syst em
1. Introduction woul d beusefulnotto objectify quality cont roloftextile
articles but al so provi de a basi st o perform onl ine
Reliabl e and accurate quality cont rol is an important inspectionforsplicejointatwi nding machine.
element in industrial textile manufacturing. For many
textileproducts,amajorquality cont rolrequirementisto In this paper, we are investigating a novel automated
judge yarn quality. Ring spinni ng produces yarn in a visualinspectionsystem fordetectionandclassifi
cationof
package form called a cop.Since copsfrom ring frames defectsencount eredinyarnopeni ng endsready forsplice.
arenotsui t
ableforfurt herprocessi ng,t hewi nding process M ethodfordetectionanddescriptionofshapeandsurface
serves to achieve additional objectives made by the defectsoftheyarnendsaredescribedandanalyzed.
requirements of the subsequent processi ng stages. The
wi nding process has t he basic funct ion of obt aini
ng a 2. The Proposed System Model for Yarn
larger package from several smal l ri ng bobbi ns. Thi s
conversi on process provi des one wi tht he possibili
ty of End Inspection
cutting outunwant edandprobl emat i
cobject ionabl efaults.
The process of removi ng such object i
onable faul t
si s 2.1Defect detection and classification
called yarn clearing.Afterremovi ng the faultsbot h yarn
endsare joined togetherusing a traditionaltechni que of The mosti mportantproperti
es oft he yarn end openi ng
knotting.Itisdi fficultto keep a high quality ofyarn by inspection arethel engt
h and unt wi strati
o oft heopeni ng
knot ti
ng,as t he knot i t
selfi s object i
onabl e due toi t s zone. Each one alone is insuffi cient for a complex
physical dimension appearance and problem during classi
fication t
ask,butt hecombi nat i
on ofbot h propert
ies
downst ream processes,t he knoti s responsible for30 t o willintroduce a good classification capability [7,8],for
60% st oppagesi nweavi ng [1,2]. that reason t he proposed syst em consi st
s of t wo
subsystems, one for unt wi
st i nspect ion and ot her for
Spl ici
ng ist he ult
imat e method t o eli
mi nate yarn faul t
s lengthi nspect
ion as shown i n Fi gure 1. In Tabl e1
andprobl emsofknot sandpi ecing.Spl ici
ng isat echni que Samplesofthedifferentgradesofyarnendsareshown.
ofjoining two yarn endsby intermingling theconstituent
fibers so thatt he jointis notsi gnifi
cantly differenti n
appearanceand mechanicalpropertieswith respectto the Untwistfeat
ure
parentyarn.Spl ici
ng technology hasgrown so rapi dlyi n
Image Preprocessi
n combiner Fi nal
recent years. M any techni ques for splicing have been classification
developed such as Electrost atic splicing, M echanical
splicing and Pneumaticsplicing.Among them,pneumatic Lengthfeat
ure
spl i
cing isthe mostpopul ar[3].The spl i
cing consi st
sof
unt wisti
ng then re-twisti
ng yarn endsusi ng airblast,i.e., Fi
gure1.Defectclassifi
cationsyst
em
firstthe yarn isopened,the fi bersintermingl ed and later

92
 Table 1. Different grades of yarn ends

Opening
Grade Opening shape
criterion
Length of the opening

optimum
zone

average

short

optimum
Untwist of the opening zone

partially

Partially overturn

Complete overturn

zero

2.2 Features for defect detection and classification Feature extraction

Preprocessing In the beginning the opening of the yarn end is checked to

reduce the time and amount of data to be processed. If the
W e implemented our algorithm in MATLAB using image- yarn end is opened, the degree of opening and the length
processing toolbox. On the beginning, we smoothed the of opening zone are checked as shown in Figure 3. In
images by standard median filtering on 5x5 pixel order to explain these as well as the overall classification
neighborhoods [9]. The filtering reduced the graininess of algorithm, consider the following:
the images and fewer false edges were detected in the next • Y: the width of parent yarn
step of segmentation. The segmentation process itself • L: the length of opening zone
eliminated many small regions of the image that resulted • The opening zone is divided to three equal areas
in noise as shown in Figure 2. As we only deal with the starting from the beginning of the opening. These
contour of the segmented image, all images were areas are :
transformed to black/white (B/ W ) ones and contours were • Area one : from the start point of the
extracted from them. opening zone to 5mm and has average
width W 1
• Area two: from 5 to 10 mm and has
(a) width W 2
• Area three: for the length greater than
10 mm with average width W 3
(b)

(c)

(d)

(e) Figure 3. Feature extraction on the image

Figure 2. Preprocessing of the image a) original

image, b) binary image, c) binary image with
noise, d) filtered image, e) B/
W contour image

93
 2.3 Classification procedure 3. Experimental Results

As presented, the procedure for automated visual In order to assess the validity of the method presented
inspection consists of two modules, one of which is used here, we have performed two sorts of experiments. One by
for untwist defect detection and classification, while the the proposed approach while the other is done by using
other is used for length detection and classification. The neural network approach.
system is based on the classification tree shown in Figure A database of 120 images (640 x 480 pixels) randomly
4 and the procedure is done as follows: collected from 12 different yarns was used to determine
the performance of the system. Images are acquired using
• If W1 is less than or equal Y, then the opening is digital camera. First, human experts analyze the images
classified as defect opening (D). and all features were extracted and described, then the
• If W1 is greater than Y and the length of the opening image processing is used. The performance of the system
zone is less than 5 mm, then the opening is classified was evaluated for classification error and classification
as short length with good opening (C1), or partially time. The definition of classification error is :
opening (C2) according to the degree of untwist
( ratio W1 / Y). & #
C .E = $ I ! × 100 (1)
• If the length is greater than 5mm and less than 10 $ T !
% "
mm, then the opening is classified as medium length
optimum opening (B1), partially opening (B2),
Where, I is the number of incorrect classification, and T is
partially overturn (E), or complete overturn (F).
the total number of images.
• If the length is greater than 10 mm , the opening is
classified as optimum opening (A) or partially
While the average time needed for classification is
overturn (E) according to the ratio between W3 and
calculated from the equation 2:
Y.
t c
= ti + t p (2)
W1/Y< 1
+ - Where: ti is the time required for acquire the image, tp is
the processing time, and tc is the total classification time.
D W1/Y!1.3
+ -
Table 2 summarizes the classification results obtained by
L!5mm L!5mm the proposed method. It can be seen from this results that,
+ - + - zero classification errors for classes A,B2,C1,C2,E and F
are obtained. While 33% and 13 % classification errors
C2 L!10mm C1 L!10m are recorded for B1 and D classes respectively. The reason
+ - + -
for that could be, some of B1 classes were classified as
W2/Y<1.2 W2/Y<1.2 W2/Y<1.2 W2/Y<1.2 C1, and some of D classes were classified as F. In general,
+ - + - + - + - the results obtained by this approach are encouraged. The
total number of incorrect classification is only three
E B2 F W3/Y<1 E B1 F W3/Y<1 samples, and the average classification error is 6%.
+ - + -

E A E A Table 2. Classification results obtained from the

proposed approach
Figure 4. classification tree Image Processing
criteria
The opening is classified as optimum or acceptable T Correct Incorrect Classification
opening if it lies above the base line as shown in Figure 5. (C) (I) error %
Otherwise, it is unacceptable opening.
A 50 50 0 0
Width ratio B1 6 4 2 33
%150 C1 B1 A
B2 8 8 0 0
%130 C2 B2
C1 4 4 0 0
Good
Base line C2 12 12 0 0
100% D
D 8 7 1 13
E Bad E 24 24 0 0
70%
F F 8 8 0 0
50%
0 5 10 length (mm) Total 120 117 3 6%
Figure 5. classification system

94
Time needed for classification is about 5 seconds on a on winding machine is presented in this paper. The
standard PC. The most time consuming portion of our presented method and algorithm were successfully tested
program by so far was the threshold and edge detection of on a limited number of yarn end samples and encouraging
the image. Translation of our MATLAB code into more results were obtained.
efficient language may enable us to reduce the
computation time. The future work will be conducted on testing of currently
developed algorithm on larger number of samples and will
4. Neural Network Approach include investigations of other methods and algorithms for
defect detection and classification.
Our second approach used detailed geometric models of
the yarn ends, and neural network. A particular open Acknowledgments
question we investigated was whether performance can be
enhanced by using neural network to recognize subtle Murata Machinery, LTD Company supported this project
differences in global features. Neural network have been by providing all database image used. The authors
successfully applied to many image classification sincerely thank the winding process team for their support
problems [10]. Neural network can run fast, can examine and helpful discussion. We would like to thank to Dr.
many competing hypotheses simultaneously, and can Morooka and members of the imaging science and
perform well with noise and distortion. We employ online engineering laboratory for helpful discussion. The first
back propagation neural Network for yarn end author has been supported by the Egyptian government
classification. Out of 120 samples, 108 samples were used scholarship for postgraduate study.
for training the network and the rest were used for the test
set. We investigate different network topologies of one or References
two dimensions and various numbers of nodes. It turns out
that one dimension [5-3-8] network architecture with
[1] Gebald G., and Leven J, Knitting advanced with the
quick propagation algorithm yields the best results. This knot free yarn specified, Knitting international, 1983-90,
network had average training correct classification rate No: 1070 February pp 57-60
96.67% . [2] Kaushik, R.C.D., Hari, P.K., Sharma, I.C., and Sarkar, A.K,
A specific illustration of the classification results is given Performance of spliced yarn in warping and weaving,
by a summary of all individual classifications. This Textile Research Journal November 1987, PP 670-673
function is illustrated in a classification confusion matrix [3] Gebald G, Remarkable quality improvements by the
in figure 6. The columns and rows of this matrix represent Autoconer splicer, the results of 10 years practical
the eight discrete grades of quality represents by the experience, W. Schlafhorst AG & CO. Blumbenberger str.
pervious approach and the net response respectively. The 143-145 D-41061 Mönchengladbach, The customer day 84,
the future of spinning
figure shows that for all 91.67% of the samples the [4] Hidetoshi Kimura, Pneumatic yarn splicing method and
classification is correct, and 8.33% classification error apparatus, USP 4,481,761 11/1984 57/22.
obtained specially for classes C2, and E. [5] Hiroshi Mima, Splicing method for spun yarns, USP
4,445,317 5/1984 57/22.
[6] Khaled Issa and Rudi Grutz, Investigation to optimize
opening process for high twisted and plied yarn, Master
thesis, Institut für Textil und Bekleidungswesen,
Niederrhein University of Applied Science,
Monchengladbach, Germany 1999.
[7] Brzakovic, D. and Vujovic, N. “Designing a defect
classification system: A case study,” Pattern recognition
29(8), pp. 1401-1419, 1996.
[8] J.livarinen and A.Visa, “An adaptive texture and shape based
defect classification,” in Proceedings of the 14th
International Conference on Pattern Recognition. (Brisbane,
Australia), Aug.16-20 1998.
[9] William K. Pratt Digital image processing, 3rd ed.,
Academic press, New York, 2001.
[10] Claus Bahlmann,Gunther Heidemann, and Helge Ritter:
“Artificial neural networks for automated quality control of
textile seams”, Pattern recognition 32(6), pp.1049-
1060,June 1999.
Figure 6. Classification confusion matrix: columns
and rows represent the eight grade of quality for the Khaled Issa, Doctoral student, Interdisciplinary Graduate
proposed grade system and the net response School of Science and Engineering, Tokyo Institute of
Technology, Japan.
Hiroshi Nagahashi, Professor, Imaging Science and
Conclusion Engineering lab, Tokyo Institute of Technology, Japan
An automated visual inspection system for defect
detection and classification of yarn ends ready for splicing

95