Using Object's Contour, Form and Depth to Embed Recognition Capability into Industrial Robots

 

I. López-Juárez^*1, M. Castelán¹, F.J. Castro-Martínez¹, M. Peña-Cabrera², R. Osorio-Comparan²

 

¹ Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional Robótica y Manufactura Avanzada Ramos Arizpe, Coahuila., México. *ismael.lopez@cinvestav.edu.mx.

² Instituto de Investigaciones en Matemáticas Aplicadas y en Sistemas Universidad Nacional Autónoma de México Mexico City, Mexico.

 

ABSTRACT

Robot vision systems can differentiate parts by pattern matching irrespective of part orientation and location. Some manufacturers offer 3D guidance systems using robust vision and laser systems so that a 3D programmed point can be repeated even if the part is moved varying its location, rotation and orientation within the working space. Despite these developments, current industrial robots are still unable to recognize objects in a robust manner; that is, to distinguish an object among equally shaped objects taking into account not only the object's contour but also its form and depth information, which is precisely the major contribution of this research. Our hypothesis establishes that it is possible to integrate a robust invariant object recognition capability into industrial robots by using image features from the object's contour (boundary object information), its form (i.e., type of curvature or topographical surface information) and depth information (from stereo disparity maps). These features can be concatenated in order to form an invariant vector descriptor which is the input to an artificial neural network (ANN) for learning and recognition purposes. In this paper we present the recognition results under different working conditions using a KUKA KR16 industrial robot, which validated our approach.

Keywords: Invariant object recognition, neural networks, shape from shading, stereo vision, robot vision.

 

RESUMEN

Los sistemas de visión para robots pueden diferenciar partes mediante el apareamiento de patrones sin considerar su orientación o localización. Algunos fabricantes ofrecen sistemas de guiado 3D utilizando sistemas robustos de visión y laser, de tal forma que un punto programado puede ser repetido aún si la parte se ha movido cambiando su orientación, localización o rotación dentro del espacio de trabajo. A pesar de estos desarrollos, los robots industriales actuales son todavía incapaces de reconocer objetos de manera robusta; esto es, distinguir un objeto de entre varios objetos similares tomando información no solo de su contorno, sino también su forma y profundidad, lo que se constituye la contribución principal de esta investigación. Nuestra hipótesis establece que es posible integrar la capacidad de reconocimiento invariante de objetos en robots industriales mediante el uso de características de contorno del objeto (información de la frontera del objeto), su forma (i.e., tipo de curvatura o información topográfica de la superficie) e información de profundidad (mediante mapas estéreo de disparidad). Estas características pueden ser concatenadas para formar un vector descriptivo que sea presentado a la entrada de una red neuronal artificial (RNA) para propósitos de aprendizaje y reconocimiento. En este artículo presentamos los resultados de reconocimiento para diferentes condiciones de trabajo empleando un robot industrial KUKA KR16, lo que valida nuestro enfoque.

 

DESCARGAR ARTÍCULO EN FORMATO PDF

 

Acknowledgements

The authors wish to thank the financial support provided by Consejo Nacional de Ciencia y Tecnología (CONACyT) through project research grant No. 61373.

 

References

[1] M. K. Hu, "Visual pattern recognition by moment invariants", IRE Trans Inform Theory. IT-8, 1962, pp. 179-187.         [ Links ]

[2] C. Yücer and K. Oflazer, "A rotation, scaling and translation invariant pattern classification system", Pattern Recognition. Vol. 26, No. 5, 1993, pp. 687-710.         [ Links ]

[3] J. Stavros and P. Lisboa, "Translation, Rotation , and Scale Invariant Pattern Recognition by High-Order Neural networks and Moment Classifiers", IEEE Transactions on Neural Networks. Vol. 3, No. 2, 1992, pp. 241-251        [ Links ]

[4] S. D. You and G. E. Ford, "Network model for invariant object recognition", Pattern Recognition Letters. Vol. 15, 1994, pp. 761-767.         [ Links ]

[5] J. Montenegro, "Hough-transform based algorithm for the automatic invariant recognition of rectangular chocolates. Detection of defective pieces", Industrial data. Vol. 9, No. 2, 2006, pp. 47-52.         [ Links ]

[6] E. González et al. "Descriptores de Fourier para identificación y posicionamiento de objetos en entornos 3D", in XXV Jornadas de Automática; Ciudad Real, Universidad de Castilla la Mancha; Sept. 2004. Available from: http://www.ceautomatica.uji.es/old/actividades/jorna das/XXV/documentos/140-liezgarthg.pdf        [ Links ]

[7] E. Bribiesca and A. Guzmán, "How to Describe Pure Form and How to Measure Differences in Shape Using Shape Numbers", PatternRecognition. Vol. 12, No. 2, 1980, pp. 101-112.         [ Links ]

[8] P. L. Worthington and E. R. Hancock, "Object recognition using shape-from shading", IEEE Trans. on Pattern Analysis and Machine Intelligence. Vol. 23, No. 5, 2001, pp. 535-542.         [ Links ]

[9] D. G. Lowe. "Distinctive Image Features from Scale-Invariant Keypoints", International Journal of Computer Vision, Vol. 60, No. 2, 2004, pp. 91-110.         [ Links ]

[10] M. Peña-Cabrera et al., "Machine Vision Approach for Robotic Assembly", Assembly Automation. Vol. 25, No. 3, 2005, pp. 204-216.         [ Links ]

[11] B. K. P. Horn. "Shape from Shading: A Method for Obtaining the Shape of a Smooth Opaque Object from One View", PhD. Thesis, MIT, 1970.         [ Links ]

[12] M. Brooks, "Two results concerning ambiguity in shape from shading," in Proceedings of the Third National Conference on Artificial Intelligence, Washington, D.C., 1983, pp.36-39.         [ Links ]

[13] R. Zhang et al., "Shape from Shading: A Survey", IEEE Trans. on Pattern Analysis and Machine Intelligence. Vol. 21, No. 8, 1999, pp. 690-706.         [ Links ]

[14] J. Koenderink and A. Van Doorn, "Surface shape and curvature scale", Image and Vision Computing. Vol. 10, 1992, pp. 557-565.         [ Links ]

[15] G. A. Carpenter and S. Grossberg, "Fuzzy ARTMAP: A neural network architecture for incremental supervised learning of analog multidimensional maps", IEEE Transactions on Neural Networks. Vol. 3, No. 5, 1992, pp. 698-713.         [ Links ]