Application of Deep Learning Approaches in Igneous Rock Hyperspectral Imaging

  • Brian Bino SinaiceEmail author
  • Youhei Kawamura
  • Jaewon Kim
  • Natsuo Okada
  • Itaru Kitahara
  • Hyongdoo Jang
Conference paper
Part of the Springer Series in Geomechanics and Geoengineering book series (SSGG)


Hyperspectral imaging has been applied in remote sensing amongst other disciplines, success in these has triggered its extensive use. Hence, it comes as no surprise that we took advantage of this technology by conducting a study aimed at the spectral analysis of several igneous rocks, and to deduce the spectral signatures of each rock unit using neural networks. Through visual observations and comparisons of these spectral signatures, parameters such as band curvature(shape), tilt(position) and strength were used for lithological discrimination. Even with this said, there often exists similarities in rocks, which are rather difficult to differentiate by means of visual or graphical analysis. However, with numerous technologies making new waves in today’s era and artificial intelligence (AI) being at the forefront of these developments, it was best fitting to employ deep learning, often referred to as a subset of AI; to train/learn from these hyperspectral signatures with a goal aimed at classifying these rocks. Deep learning has networks such as the convolution neural network (CNN), which has algorithms that excel in feature representation from visual imagery; taking into account that the more data is fed into the training process and later used as a database for further training, the higher the future prediction accuracy. Gathered outcomes from the CNN show exceptionally high prediction accuracy capabilities of 96%; suggesting viable field and laboratory usage of these systems as a unit for mining and rock engineering applications.


Deep learning Hyperspectral imaging Convolution neural network 


  1. Jia, F., Lein, Y., Lin, J., Zhou, X., Lu, N.: Deep neural networks: a promising tool for fault characteristic mining and intelligent diagnosis of rotating machinery with massive data. Mech. Syst. Signal Process. 72–73, 303–315 (2016)CrossRefGoogle Scholar
  2. Xing, J., Li, K., Hu, W., Yuan, C., Ling, H.: Diagnosing deep learning models for high accuracy age estimation from a single image. Pattern Recogn. 66, 106–116 (2017)CrossRefGoogle Scholar
  3. Simonyan, K., Vedaldi, A., Zisserman, A.: Deep inside convolutional networks: visualising image classification models and saliency maps (2013)Google Scholar
  4. Kruse, F.A.: Mapping surface mineralogy using imaging spectrometry. Geomorphology 137, 41–56 (2012)CrossRefGoogle Scholar
  5. Meer, F.: The effectiveness of spectral similarity measures for the analysis of hyperspectral imagery. Int. J. Appl. Earth Obs. Geoinf. 8, 3–17 (2006)CrossRefGoogle Scholar
  6. Pieters, C.M., et al.: The Moon Mineralogy Mapper (M3) on Chandrayaan-1. Curr. Sci. 96(4), 500–505 (2009)Google Scholar
  7. Poria, S., Cambria, E., Gelbukh, A.: Aspect extraction for opinion mining with a deep convolutional neural network. Knowl.-Based Syst. 108, 42–49 (2016)CrossRefGoogle Scholar
  8. Tompkins, S., Pieters, C.M.: Mineralogy of the lunar crust: results from Clementine. Meteor. Planet. Sci. 34, 25–41 (1999)CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  • Brian Bino Sinaice
    • 1
    Email author
  • Youhei Kawamura
    • 1
  • Jaewon Kim
    • 1
  • Natsuo Okada
    • 2
  • Itaru Kitahara
    • 3
  • Hyongdoo Jang
    • 4
  1. 1.Graduate School of International Resource SciencesAkita UniversityAkitaJapan
  2. 2.Faculty of International Resource SciencesAkita UniversityAkitaJapan
  3. 3.Center for Computational SciencesUniversity of TsukubaTsukubaJapan
  4. 4.Western Australian School of Mines Minerals, Energy and Chemical EngineeringCurtin UniversityKalgoorlieAustralia

Personalised recommendations