Abstract
A higher-order neural network (HONN) can be designed to be invariant to geometric transformations such as scale, translation, and in-plane rotation. Invariances are built directly into the architecture of a HONN and do not need to be learned. Thus, for 2D object recognition, the network needs to be trained on just one view of each object class, not numerous scaled, translated, and rotated views. Because the 2D object recognition task is a component of the 3D object recognition task, built-in 2D invariance also decreases the size of the training set required for 3D object recognition. We present results for 2D object recognition both in simulation and within a robotic vision experiment and for 3D object recognition in simulation. We also compare our method to other approaches and show that HONNs have distinct advantages for position, scale, and rotation-invariant object recognition.
The major drawback of HONNs is that the size of the input field is limited due to the memory required for the large number of interconnections in a fully connected network. We present partial connectivity strategies and a coarse-coding technique for overcoming this limitation and increasing the input field to that required by practical object recognition problems.
Article PDF
Similar content being viewed by others
Explore related subjects
Discover the latest articles, news and stories from top researchers in related subjects.Avoid common mistakes on your manuscript.
References
Casasent, D., & Chang, W.-T. (1985.) Parameter estimation and in-plane distortion invariant chord processing. SPIE,579, 2–10.
Chen, Z., & Ho, S.-Y. (1986.) Computer vision for robust 3D aircraft recognition with fast library search.Pattern Recognition, 24, 375–390.
Giles, G.L., & Maxwell, T. (1987.) Learning, invariances, and generalization in high-order neural networks.Applied Optics, 26, 4972–4978.
Giles, G.L., Griffin, R.D., & Maxwell, T. (1988.) Encoding geometric invariances in higher-order neural networks.Neural Information Processing Systems, American Institute of Physics Conference Proceedings, (pp. 301–309).
Haberman, R. (1983.)Elementary applied partial differential equations. Englewood Cliffs, NJ: Prentice-Hall.
Hsu, Y.-N., Arsenault, H.H., & April, G. (1982.) Rotation-invariant digital pattern recognition using circular harmonic expansion.Applied Optics, 21/22, 4012–4015.
Hu, M. (1962.) Visual pattern recognition by moment invariants.IRE Transactions on Information Theory, IT-8, 179–187.
Jared, D.A., & Ennis, D.E. (1989.) Inclusion of filter modulation in synthetic discriminant-function construction.Applied Optics, 28, 232–239.
Kuhl, F.P., & Giardina, C.R. (1982.) Elliptic Fourier feature of a closed contour.Computer Vision, Graphics, and Image Processing, 18, 236–258.
Pitts, W., & McCulloch, W.S. (1947.) How we know universals: The perception of auditory and visual forms.Bulletin of Mathematical Biophysics, Chicago: University of Chicago Press,9, 127–147.
Quinlan, J.R. (1986.) Induction of decision trees.Machine Learning, 1, 81–106.
Reid, M.B., Spirkovska, L., & Ochoa, E. (1989.) Rapid training of higher-order neural networks for invariant pattern recognition.Proceedings of Joint International Conference on Neural Networks (Vol. 1, pp. 689–692), Washington, D.C.
Reid, M.B., Ma, P.W., Downie, J.D., & Ochoa, E. (1990a.) Experimental verification of modified synthetic discriminant function filters for rotation invariance.Applied Optics, 29, 1209–1214.
Reid, M.B., Ma, P.W., & Downie, J.D. (1990b). Determining object orientation with a hierarchical database of binary synthetic discriminant function filters.Japanese Journal of Applied Physics, 29, 1284–1286.
Rosenfeld, R., & Touretzky, D.S. (1988.) A survey of coarse-coded symbol memories.Proceedings of the 1988 Connectionist Models Summer School, Carnegie Mellon University (pp. 256–264).
Rumelhart, D.E., Hinton, G.E., & Williams, R.J. (1986.) Learning internal representations by error propagation. InParallel Distributed Processing (Vol. 1). Cambridge, MA: MIT Press.
Spirkovska, L., & Reid, M.B. (1992.) Application of higher-order neural networks in the PSRI object recognition domain. In B. Soucek and the IRIS Group (Eds.),Fuzzy, holographic, invariant and parallel intelligence: The sixth generation breakthrough. New York: Wiley.
Spirkovska, L., & Reid, M.B. (1990.) An empirical comparison of ID3 and HONNs for distortion invariant object recognition.Proceedings of the Second International Conference on Tools for Artificial Intelligence (pp. 577–582). Washington, D.C.
Sullins, J. (1985.) Value cell encoding strategies (Technical report TR-165). Computer Science Department, University of Rochester, Rochester, NY.
Troxel, S.E., Rogers, S.K., & Kabrisky, M. (1988.) The use of neural networks in PSRI recognition.Proceedings of Joint International Conference on Neural Networks (pp. 593–600). San Diego, CA.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Spirkovska, L., Reid, M.B. Higher-order neural networks applied to 2D and 3D object recognition. Mach Learn 15, 169–199 (1994). https://doi.org/10.1007/BF00993276
Received:
Accepted:
Issue Date:
DOI: https://doi.org/10.1007/BF00993276