Classification of Arsenic Bearing Minerals Using Hyperspectral Imaging and Deep Learning for Mineral Processing

DOI Web Site Web Site 7 References
  • OKADA Natsuo
    Student, Graduate School of International Resources Sciences, Akita University
  • MAEKAWA Yohei
    Student, Department of Transdisciplinary Science and Engineering, School of Environment and Society, Tokyo Institute of Technology
  • OWADA Narihiro
    Student, Faculty of International Resources Sciences, Akita University
  • HAGA Kazutoshi
    Associate Professor, Graduate School of International Resources Sciences, Akita University
  • SHIBAYAMA Atsushi
    Professor, Graduate School of International Resources Sciences, Akita University
  • KAWAMURA Youhei
    Professor, Graduate School of International Resources Sciences, Akita University

Bibliographic Information

Other Title
  • 選鉱プロセスにおけるハイパースペクトルイメージングと深層学習を用いたヒ素含有鉱石の分類
  • センコウ プロセス ニ オケル ハイパースペクトルイメージング ト シンソウ ガクシュウ オ モチイタ ヒソ ガンユウ コウセキ ノ ブンルイ

Search this article

Description

<p>Currently, there have been issues concerning the depletion and scarcity of mineral resources. This is mostly due to the excavation of high grade minerals having already occurred years and years ago, hence forcing the mining industry to opt for the production and optimization of lower grade minerals. This however brings about a plethora of problems, many of which economic, stemming from the purification of those low grade minerals in various stages required for mineral processing. In order to reduce costs and aid in the optimization of the mining stream, this study, introduces an automatic mineral identification system which combines the predictive abilities of deep learning with the excellent resolution of hyperspectral imaging, for pre-stage of mineral processing. These technologies were used to identify and classify high grade arsenic (As) bearing minerals from their low grade mineral counterparts non-destructively. Most of this ability to perform such tasks comes from the highly versatile machine learning model which employs deep learning as a means to classify minerals for mineral processing. Experimental results supported this statement as the model was able to achieve an over 90% accuracy in the prediction of As-bearing minerals, hence, one could conclude that this system has the potential to be employed in the mining industry as it achieves modern day system requirements such as high accuracy, speed, economic, userfriendly and automatic mineral identification.</p>

Journal

  • Journal of MMIJ

    Journal of MMIJ 137 (1), 1-9, 2021-01-31

    The Mining and Materials Processing Institute of Japan

References(7)*help

See more

Keywords

Details 詳細情報について

Report a problem

Back to top