Détection de défauts temps réel sur des objets à géométrie complexe: étude par svm, boosting et hyperrectangles

Détection de défauts temps réel sur des objets à géométrie complexe: étude par svm, boosting et hyperrectangles

Real-time flaw detection on complex part : study of svm, boosting and hyperrectangle based method

S. Bouillant J. Mitéran  M. Paindavoine  J. Matas 

Le2i, Aile des Sciences de l’ingénieur, Université de Bourgogne, BP 47870 21078 Dijon

Center for Machine Perception, CVUT, Karlovo Namesti 13, Prague, République Tchèque

Corresponding Author Email: 
miteranj@u-bourgogne.fr
Page: 
55-69
|
Received: 
25 April 2003
|
Accepted: 
N/A
|
Published: 
29 February 2004
| Citation

OPEN ACCESS

Abstract: 

We present in this paper our works on the classification of industrial parts based on «Support Vector Machine» method. We present the practical frame in which are made the operations, flaws types to detect as well as feature extraction techniques. Then we introduce the three classification techniques we implemented. We explain our learning method and how we obtain optimum classifier parameters. We compare the results obtained using feature space based on a priori knowledge and on space extracted from sequential selection algorithm.

Résumé

Nous présentons dans cet article une application complète des « Support Vector Machine» au contrôle qualité par vision artificielle de pièces à géométrie complexe. Nous précisons le cadre pratique dans lequel s’effectuent les opérations, la nature des défauts à détecter ainsi que les techniques d’extraction des paramètres discriminants. Nous présentons ensuite les trois méthodes de classification utilisées. Nous définissons le protocole d’apprentissage, ainsi que la méthode de recherche des paramètres optimum du classifieur. Nous comparons les résultats obtenus à partir d’un espace de description défini a priori ainsi que ceux issus d’une sélection de paramètres via un algorithme séquentiel.

Keywords: 

Classification, SVM, Hyperrectangle, Boosting, quality control, image processing, pattern analysis

Mots clés

Classification, SVM, Hyperrectangle, Boosting, analyse de formes, contrôle qualité, image

1. Introduction
2. Présentation Du Problème Et Extraction Des Paramètres
3. Méthodes De Classification Étudiées
4. Recherche Des Paramètres De Classification Optimaux
5. Conclusion Et Perspectives
  References

[1] M.M.Ratnam and C.P.Lim, «Classification using shadow moiré and neural network» in Proc. of QCAV 2001 – Vol. 2, pp. 355-358, 2001.

[2] C.M.Bishop, Neural network for pattern recognition, Oxford university press, 1995.

[3] C.Firmin, D.Hamad, J.-G.Postaire and R.D.Zhuang, «Gaussian neural network for glass bottle production: a learning procedure » in International Journal on Neural Systems – Vol. 8, pp. 41-60, 1996.

[4] B.Dubuisson, Diagnostic, intelligence artificielle et reconnaissance des formes (Traité IC2 série productique), 2001.

[5] R.O.Duda, P.E.Hart and D.G.Stork, Pattern classification, John Wiley & Son, INC, 2001.

[6] S.Salzberg, «A nearest hyperrectangle learning method» in Machine Learning – Vol. 6, pp. 251-276, 1991.

[7] J.Miteran, P.Gorria and M.Robert, «Classification géométrique par polytopes de contraintes. Performances et intégration» in Traitement du Signal – Vol. 11(:5), pp. 393-408, 1994.

[8] Y.Freund, «Boosting a weak learning algorithm by majority » in Information and Computation – Vol. 121(2), pp. 256-285, 1995.

[9] R.E.Schapire, «A brief introduction to boosting » in Proc. of 16th International Joint Conference on Artificial Intelligence, pp. 1401-1405, 1999.

[10] V.N.Vapnik, The nature of statistical learning theory, SpringerVerlag, New-York, 1995.

[11] J. Miteran, J-P. Zimmer, F. Yang and M. Paindavoine, « Access control: adaptation and real time implantation of a face recognition method » in Optical Engineering – Vol. 40(4), pp. 586-593, 1-4-2001.

[12] P.Geveaux, S.Kohler, J.Miteran, F.Truchetet and F.Meriodaudeau, «Comparison between two classification methods, Application to defects detection by artificial vision in industrial field» in Proc. of SPIE – Vol. 3966A, pp. 154-161, 1-1-2000.

[13] J.Miteran, S.Bouillant and E.Bourennane, «SVM approximation for real-time image segmentation by using an improved Hyperrectanglesbased methods » in Real Time Imaging, Academic Press – Vol. 9(3), pp. 179-188, 2003.

[14] J.Miteran, P.Geveaux, R.Bailly and P.Gorria, «Real time defect detection using image segmentation » in Proc. of ISIE 97, pp. 713-716, 1-7-1997.

[15] W.H.Tsai, «Moment preserving thresolding: a new approach » in Computer vision, graphics and image processing – Vol. 29, pp. 377-393, 1-12-1985.

[16] G.J. McLachlan, Discriminant Analysis and Statistical Pattern Recognition, John Wiley & Sons, New York, 1992.

[17] R.A.Fisher, «The use of multiple measurements in taxonimic problem » in Annales of Eugenics Vol. 7, Part II, pp. 179-188, 1936.

[18] B.Schölkopf, A.Smola, K.R.Müller, C.J.C.Burges and V.N.Vapnik, «Support vector methods in learning and feature extraction» in Australian Journal of Intelligent Information Processing Systems – Vol. 1(1), pp. 3-9, 1998.

[19] Jonsson K., Smola A., Mullër K.R. and Burges C., «Support Vector Machines for Face Authentication» in Proc. of British Machine Vision Conference, pp. 543-553, 1999.

[20] M.A.Hearst, B.Schölkopf, S.Dumais, E.Osuna and J.Platt, «Trens and Controversies – Support Vector Machines » in IEEE Intelligent systems – Vol. 13(4), pp. 18-28, 1998.

[21] S.Canu, «Modèles connexionnistes et machines à vecteurs supports pour la décision » in Décision de reconnaissance des formes en Signal, Hermes, 2002, pp. 181-217.

[22] K.Tieu and P.Viola, «Boosting Image Retrieval » in International Journal of Computer Vision – 561-3, pp. 17-36, 2004.

[23] R.E.Schapire and Y.Singer, «A boosting-based system for text categorization» in Machine Learning – 392/3, pp. 135-168, 2000.

[24] D.Wettschereck and T.Dietterich, «An Experimental Comparison of the Nearest-Neighbor and Nearest-Hyperrectangle Algorithms» in Machine Learning – Vol. 19(6), pp. 5-27, 1995.

[25] J.Miteran, R.Bailly and P.Gorria, «Classification board for real time image segmentation » in Proc. of ICASSP 97, pp. 4069-4072, 1-4-1997.

[26] V.N.Vapnik, Statistical learning theory, John Wiley & sons, INC, 1998.

[27] R.Fletcher, Pratical methods of optimisation, John Wiley & Son, INC, 1987.

[28] B.E.Boser, I.M.Guyon and V.N.Vapnik, «A training algorithm for optimal margin classifiers» in Proc. of Fifth annual workshop on computional learning theory, pp. 144-152, 1992.

[29] R.O.Duda, P.E.Hart and D.G.Stork, Chap. n°5 «Linear Discriminant functions » in Pattern classification, John Wiley & Sons, 2001, pp. 264-265.

[30] R.Courant and D.Hilbert, Methods of mathematical physics, InterScience, New York, 1953.

[31] S.Gunn, MatLab SVM toolbox, http://www.isis.ecs.soton.ac.uk/resources/svminfo/download.php – v 1.0 – 2001.

[32] C.C.Chang and C.J.Li, LIBSVM – A library for support vector machines, http://www.csie.ntu.edu.tw/~cjlin/libsvm/index.html – v. 2.33 – 2002.

[33] C.M.Bishop, Chap. n°9 «Learning and generalization» in Neural network for pattern recognition, Oxford university press, 1995, pp. 372-375.

[34] S.Bouillant, J.Miteran, F.Yang and M.Paindavoine, «Real-time flaw detection on complex part: study of SVM and Hyperrectangle based method » in Proc. of ICASSP 2002, pp. 3596-3599, 1-1-2002.

[35] V.N.Vapnik, Chap. n°12 «SV machines for pattern recognition » in Statistical learning theory, John Wiley & Sons, 1998, pp. 496-498.

[36] M. Dash and H. Liu, « Feature selection for classification » in Intelligent data analysis – Vol. 1(3), pp. 131-156, 1997.

[37] A.Jain and D.Zongker, «Feature selection: evaluation, application, and small sample performance» in IEEE Transactions on pattern analysis and machine intelligence – Vol. 19(2), pp. 153-158, 1997.

[38] D.E. Goldberg, Genetic algorithms in search, optimization and machine learning, Addison Wesley, 1989.

[39] J.R.Koza, Genetic Programming, MIT Press, 1992.

[40] F.Ferri, P.Pudil, M.Hatef and J.Kittler, «Comparative study of techniques for large-scale feature selection» in Pattern Recognition in Practice IV, pp. 403-413, 1994.

[41] J.Kittler, «Feature set algorithms» in Pattern recognition and signal processing, pp. 41-60, 1978.

[42] P.Somol, P. Pudil, J.Novovicova and P.Paclik, «Adaptative floating search methods in feature selection » in Pattern recognition letters – Vol. 10, pp. 335-347, 1989.