SciELO - Scientific Electronic Library Online

 
vol.9 issue1Theoretic-Experimental Analysis of Comparison Criteria for Object-Oriented Conceptual SchemasImplementation of a Neural Hierarchical Multimodel for Identification and Control of Mechanical Systems author indexsubject indexsearch form
Home Pagealphabetic serial listing  

Services on Demand

Journal

Article

Indicators

Related links

  • Have no similar articlesSimilars in SciELO

Share


Computación y Sistemas

Print version ISSN 1405-5546

Comp. y Sist. vol.9 n.1 México Jul./Sep. 2005

 

Clasificación de Defectos en Madera utilizando Redes Neurales Artificiales

 

Wood Defects Classification Using Artificial Neural Network

 

Graciela María de Jesús Ramírez Alonso y Mario Ignacio Chacón Murguía

 

Instituto Tecnológico de Chihuahua Laboratorio de DSP y Visión Av. Tecnológico 2909, Chihuahua, Chih., México C.P. 31310 Tel.4–13–74–74 Ext 112 y 114 gramirez@itchihuahua.edu.mx ; mchacon@itchihuahua.edu.mx

 

Artículo recibido en junio 06, 2004
Aceptado en abril 01, 2005

 

Resumen

Este artículo describe un clasificador neural que diferencía entre 7 tipos de defectos en maderas llamados botones. La inspección visual de estos defectos por humanos tiene un alto grado de complejidad debido a la varianza intraclase. Las características utilizadas se extrajeron de las imágenes de maderas mediante filtros Gabor de 2D. Estos filtros son pasa banda selectivos a la orientación y frecuencia, muy utilizados para imágenes en donde la textura es un factor importante. Para optimizar las características se realizó una reducción de dimensión del resultado de los filtros Gabor mediante el método de Análisis de Componentes Principales. La red neural que se implementó fue una red Perceptrón multicapa de 3 capas entrenada con el algoritmo de Resalient Backpropagation. La tasa de reconocimiento de la red fue de un 83.91%, siendo este resultado aceptable teniendo en cuenta que un inspector humano alcanza un reconocimiento entre el 75 y 85%.

Palabras Clave: Redes neurales, filtros Gabor, procesamiento de imágenes.

 

Abstract

This paper describes a neural classifier to classify 7 different wood defects called knots. Human visual inspection of these defects involves a high degree of complexity due to inter–class variance. 2D Gabor filters were used for feature extraction. These filters are selective band pass filters to orientation and frequency. These filters are used where texture is an important feature. The method of principal component analysis was used to reduce the number of features generated by the Gabor filters. The neural network implemented was a multilyer perceptron with 3 layers trained with the Resalient backpropagation algorithm. The performance of the classifier was 83.91% of correct classification. This result is acceptable considering that the performance of a human inspector is 75% to 85%.

Keywords: Neural networks, Gabor filters, image processing.

 

DESCARGAR ARTÍCULO EN FORMATO PDF

 

Agradecimientos

Los autores agradecen a CONACYT por el financiamiento de este trabajo bajo el convenio PFPN–03–29–05.

 

Referencias

1. Haykin S.: Neural Networks A comprehensive foundation. Prentice Hall (1999).        [ Links ]

2. Lin C. y George Lee S.C.: Neural Fuzzy Systems. A Neuro–Fuzzy Synergism to Intelligent Systems. Prentice Hall (1996).        [ Links ]

3. Kohonen T.: Self Organizing Maps. Springer–Verlag, Berlin, Germany (1995).        [ Links ]

4. Silven O., Niskanen M. and Kauppinen H.: Wood inspection with non–supervised clustering. Machine Vision and application. Vol. 13 (2003) 275–285.        [ Links ]

5. Silven O., Niskanen M. and Kauppinen H. University of Oulu. Department of electrical engineering. URL: http://www.oulu.fi/research/imag/gsom/.         [ Links ]

6. Lampinen J. and Smolander S.: Word defect recognition: A comparative study. Proc. of workshop on Machine vision in advanced production. Oulu, Finland (1994).        [ Links ]

7. Lampinen J., Smolander S. and Korhonen M.: Wood surface inspection system based on generics visual features. International Conference on artificial neural networks ICANN'95, Paris, October 9–13(1995).        [ Links ]

8. Lampinen J. and Smolander S.: Self organizing feature extraction in recognition of wood surface defects and color images. International Journal of Pattern Recognition and Artificial intelligence. Vol. 10, No. 2 (1996) 97–113.        [ Links ]

9. Kauppinen H., Silven O. and Piirainen T.: Self Organizing Map based user interface for visual surface inspection. Proc. 11th Scandinavian conference on image analysis. Kangerlussuaq, Greenland, June 7–11 (1999) 801–808.        [ Links ]

10. Aznar–Casanova J. A.: Análisis multiescala y multiorientación de imágenes mediante un banco de filtros de Gabor–2D. Revista Cognitiva; Vol. 12, no.2, 2002, pp 223–246.        [ Links ]

11. Liu C. and Wechsler H.: Independent component analysis of Gabor features for face recognition. IEEE Trans. Neural Networks. Vol. 14 (2003) 919–928.        [ Links ]

12. Tsoukalas L. H., Uhrig R. E..: Fuzzy and Neural Approaches in Enginnering, Wiley Interscience (1997).        [ Links ]

13. Somaie A.A., Badr A., Salah, T.: Aircraft image recognition using back–propagation. Radar, CIE International Conference on, Proceedings, 15–18 Oct (2001) 498–501.        [ Links ]

14. Riedmiller M., Braun H.: A direct adaptive method for faster backpropagation learning: the RPROP algorithm. Neural Networks, IEEE International Conference on. Vol. 1, 28 March –1 April (1993) 586–591.        [ Links ]

15. Chacón M. M., Valenzuela I.: Fast image processing application development scheme for the DSK C6711 using Matlab and Simulink. IEEE Workshop on Signal Processing Education, Proceedings, Ago. (2004), 79–83.        [ Links ]

Creative Commons License All the contents of this journal, except where otherwise noted, is licensed under a Creative Commons Attribution License