• No results found

A Database Optical Device microscopic Number of

fields

Corresponding

area Total items number Class 1 items number Class -1 items number

A 1 1178 28 cm² 3865 275 3590

B 2 605 14 cm² 1910 184 1726

Table 1. Description of the two databases used for validation experiments

class 1 classification and class -1 classification perfor-

mances. It shows first that equivalent performances can

be obtained using only 5-dimensional data instead of unprocessed defects representations (13-dimensional). As a consequence neural architecture complexity and therefore processing time can be saved using CDA dimensionality reduction, while keeping performance level. Moreover, obtained scores are satisfactory: about

70% of “dust” defects are well-recognized (this can be

enough for aimed application) as well as about 97% of other defects (the few 3% errors can however pose

problems because every “permanent” defect has to be reported). Furthermore, we think that this significant

performances difference between class 1 and class -1 recognition is due to the fact that class 1 is underrep- resented in learning database.

Figure 4. Classification performances for different CDA issued data dimensionality. Classification performances using raw data (13-dimensional) are also depicted as dotted lines.

50 55 60 65 70 75 80 85 90 95 100 2 3 4 5 6 7 8 9 10 11 12 13 Data Dimensionality % o f c la s s if ie r c o rr e c t a n s w e rs

Class -1 c lassification pe rform ance Class 1 classification per form ance Global classification per form ance

FUTURE TRENDS

Next phase of this work will deal with classification

tasks involving more classes. We want also use much more Fourier-Mellin invariants, because we think

that it would improve classification performance by

supplying additional information. In this case, CDA based dimensionality reduction technique would be a

foremost step to keep reasonable classification system’s

complexity and processing time.

CONCLUSION

A reliable diagnosis of aesthetic flaws in high-quality

optical devices is a crucial task to ensure products’

nominal specification and to enhance the production

quality by studying the impact of the process on such defects. To ensure a reliable diagnosis, an automatic system is needed to detect defects and secondly dis-

ANN-Based Defects’ Diagnosis of Industrial Optical Devices

criminate the “false” defects (correctable defects) from “abiding” (permanent) ones. In this paper is described

a complete framework, which allows detecting all de- fects present in a raw Nomarski image and extracting

pertinent features for classification of these defects. Obtained proper performances for “dust” versus “other” defects classification task with MLP neural network has

demonstrated the pertinence of proposed approach. In addition, data dimensionality reduction permits to use

low complexity classifier (while keeping performance

level) and therefore to save processing time.

REFERENCES

Bouchareine, P. (1999). Métrologie des Surfaces.

Techniques de l’Ingénieur, R1390.

Chatterjee, S. (2003). Design Considerations and Fabri-

cation Techniques of Nomarski Reflection Microscope.

Optical Engineering, 42, 2202-2212.

Choksuriwong, A., Laurent, H., & Emile, B. (2005). Comparison of invariant descriptors for object rec- ognition. IEEE International Conference on Image Processing (ICIP), 377-380.

Demartines, P. (1994). Analyse de Données par Réseaux de Neurones Auto-Organisés. PhD Thesis - Institut National Polytechnique de Grenoble.

Demartines, P. & Hérault, J. (1993). Vector Quantiza- tion and Projection Neural Network. Lecture Notes in Computer Science, 686, 328-333.

Demartines, P. & Hérault, J. (1995). CCA : “Curvilinear

Component Analysis”. Proceedings of 15th workshop GRETSI.

Derrode, S. (1999). Représentation de Formes Planes à Niveaux de Gris par Différentes Approximations de Fourier-Mellin Analytique en vue d’Indexation de Bases d’Images. PhD Thesis - Université de Rennes I. Egmont-Petersen, M., de Ridder, D., & Handels, H. (2002). Image Processing with Neural Networks - A Review. Pattern Recognition, 35, 2279-2301.

Flewitt, P. E. J. & Wild, R. K. (1994). Light Microscopy. In Physical Methods for materials characterisation.

Ghorbel, F. (1994). A Complete Invariant Description for Gray Level Images by the Harmonic Analysis Ap- proach. Pattern Recognition, 15, 1043-1051.

Lee, J. A., Lendasse, A., Donckers, N., & Verleysen, M. (2000). A Robust Nonlinear Projection Method. In European Symposium on Artificial Neural Networks - ESANN’2000.

Lennon, M., Mercier, G., Mouchot, M. C., & Hubert- Moy, L. (2001). Curvilinear Component Analysis for Nonlinear Dimensionality Reduction of Hyperspectral Images. Proceedings of SPIE, 4541, 157-168.

Verleysen, M. (2001). Learning high-dimensional data. In LFTNC’2001 - NATO Advanced Research Workshop on Limitations and Future Trends in Neural Computing.

Voiry, M., Houbre, F., Amarger, V., & Madani, K. (2005). Toward Surface Imperfections Diagnosis Using Optical Microscopy Imaging in Industrial Environment. IAR & ACD Workshop 2005 Proceedings, 139-144.

Voiry, M., Madani, K., Amarger, V., & Bernier, J. (2007). Impact of Data Dimensionality Reduction on Neural

Based Classification: Application to Industrial Defects Classification. Proceedings of the 3rd International Workshop on Artificial Neural Networks and Intelligent Information Processing - ANNIIP 2007, 56-65. Voiry, M., Madani, K., Amarger, V., & Houbre, F. (2006). Toward Automatic Defects Clustering in Industrial Production Process Combining Optical Detection and

Unsupervised Artificial Neural Network Techniques.

Proceedings of the 2nd International Workshop on Artificial Neural Networks and Intelligent Information Processing - ANNIIP 2006, 25-34.

Zhang, G. P. (2000). Neural Networks for Classifica- tion: A Survey. IEEE Trans.on Systems, Man, and Cybernetics - Part C: Applications and Reviews, 30,

451-462.

KEy TERMS

Artificial Neural Networks: A network of many

simple processors (“units” or “neurons”) that imitates

a biological neural network. The units are connected by unidirectional communication channels, which carry numeric data. Neural networks can be trained

ANN-Based Defects’ Diagnosis of Industrial Optical Devices

A