Abstract
All the pixels of an image do not contain the same amount of information. Typically, the borders of an image contain less information than the centre. This paper introduces a methodology to locate the most relevant regions in images. A relevant region is a set of pixels that contains suitable information to recognise an image class versus the rest. To perform a detailed analysis of images, they are divided into regions and for each region a conformal predictor is trained. The values of the non-conformity measure are used, on the one hand, to hedge the classifier outputs with confidence and credibility measures and, on the other hand, to choose the most relevant regions. The combination of the best regions and their predictors (one per each class) is used to classify new images. The dimensionality of the original images is reduced to the dimension of the regions combination. This technique has been applied to the classification of images in the Thomson scattering diagnostic of a nuclear fusion device: the TJ-II stellarator. There are five different types of images. A database of more than 1200 TJ-II Thomson scattering images has been analyzed.
Article PDF
Similar content being viewed by others
Rferences
Gammerman A., Vovk V.: Hedging predictions in machine learning. Comput. J. 50(2), 151–163 (2007). doi:10.1093/comjnl/bxl065
Barth C.J., Pijper F.J., Meiden H.J.v.d., Herranz J., Pastor I.: “High-resolution multiposition Thomson Scattering for the TJ-II stellarator”. Rev. Sci. Instrum. 70(1), 763–767 (1999)
Herranz J., Castejón F., Pastor I., McCarthy K.J.: The spectrometer of the high-resolution multiposition Thomson scattering diagnostic for TJ-II. Fusion Eng. Design. 64(4), 525–536 (2003)
Sanchez, J., et al.: Overview of TJ-II experiments. Nuclear Fusion 51, 094022 (2011) (10 pp). doi:10.1088/0029-5515/51/9/094022
Vega, J., et al.: Application of intelligent classification techniques to the TJ-II Thomson Scattering diagnostic. In: Proceedings of the 32nd EPS conference on plasma Phys. ECA vol. 29C, P-2.090. Tarragona, 27 June-1 July (2005)
Makili, L., Vega, J., Dormido-Canto, S., Pastor, I., Pereira, A., Farias, G., Portas, A., Perez-Risco, D., Rodriguez-Fernandez, M.C., Busch, P.: Upgrade of the automatic analysis system in the TJ-II Thomson Scattering diagnostic: New image recognition classifier and fault condition detection. Fusion Engineering and Design, vol. 85, issues 3–4, Proceedings of the 7th IAEA Technical Meeting on Control, Data Acquisition, and Remote Participation for Fusion Research, pp. 415–418, July 2010, ISSN: 0920-3796. doi:10.1016/j.fusengdes.2009.10.004
Farias G., Dormido R., Santos M., Duro N.: Image classifier for the TJ-II Thomson scattering diagnostic: evaluation with a feed forward neural network. Lect. Notes Comput. Sci. 3562, 362–381 (2005). doi:10.1007/11499305_62
Vega J., Murari A., Pereira A., Gonzalez S., Pastor I.: Accurate and reliable image classification by using conformal predictors in the TJ-II Thomson scattering. Rev. Sci. Instrum. 81, 10E118 (2010). doi:10.1063/1.3478689
Makili, L., Vega, J., Dormido-Canto, S., Pastor, I., Murari, A.: Computationally efficient SVM multi-class image recognition with confidence measures. Fusion Eng. Design, corrected proof, Available online 22 March 2011, ISSN: 0920-3796 (in press). doi:10.1016/j.fusengdes.2011.02.081
Daubechies, I.: Ten lectures on wavelets. CBMS-NSF conference series in applied mathematics. SIAM Ed (1992)
Saunders, C., Gammerman, A., Vovk, V.: Transduction with confidence and credibility. Proc. IJCAI’99 2, 722–726 (1999)
Vapnik, V.: Statistical learning theory. Wiley, INC.m, London (1998)
Papadopoulos, H.: Inductive Conformal Prediction: Theory and Application to Neural Networks. Tools in Artificial Intelligence, pp. 315–330. I-Tech, Vienna (2008)
Shafer G., Vovk V.: A tutorial on conformal prediction. J. Mach. Learn. Res. 9, 371–421 (2008)
Papadopoulos, H., Proedrou, K., Vovk, V., Gammerman, A.: Inductive confidence machines for regression. In: Proceedings of the 13th European Conference on Machine Learning (ECML’02). Lecture Notes in Computer Science, vol. 2430, pp. 345–356. Springer, Berlin (2002)
Weston, J., Watkins, C.: Support vector machines for multi-class pattern recognition. ESANN Proc. 219-224 (1999), ISBN: 2-600049-9-X
Mallat, S.: A wavelet tour of signal processing, 2nd edn. Academic Press, New York (2001)
Cherkassky, V., Mulier, F.: Learning from data: concepts, theory and methods, 2nd edn. Wiley, New York (2007)
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
González, S., Vega, J., Pereira, A. et al. Region selection and image classification methodology using a non-conformity measure. Prog Artif Intell 1, 215–222 (2012). https://doi.org/10.1007/s13748-012-0020-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13748-012-0020-z