A system of the granite weathering degree assessment using hyperspectral image and CNN. Issue 5 (28th May 2022)

• Record Type: Journal Article
• Authors: Kim, Jaewon; Kawamura, Youhei; Nishikawa, Osamu; Sinaice, Brian Bino; Okada, Natsuo; Utsuki, Shinji
• Is Part Of: International journal of mining, reclamation and environment. Volume 36:Issue 5(2022)
• Page Start: 368
• Page End: 380
• Publication Date: 2022-05-28
• Subjects: Hyperspectral imaging -- convolutional neural networks -- weathering degree of rock -- granite; Mining engineering -- Periodicals; Mineral industries -- Environmental aspects -- Periodicals; Abandoned mined lands reclamation -- Periodicals; 622.292
• DOI: 10.1080/17480930.2022.2051135
• Languages: English
• ISSNs: 1748-0930

Abstract

ABSTRACT: Due to weathering processes that occur all around the world, rocks that are exposed eventually undergo physical and chemical changes. These weathering processes have the potential to turn fresh rocks into soft and weakened derivatives, leading to the collapse and subsidence of structures, such as tunnels, dams, and others, as well as natural structures, such as cliffs and slopes. This study attempts to employ hyperspectral imaging and a deep learning network called Convolutional Neural Network (CNN) in the effort to achieve a reliable process for rock weathering degree assessment. This study employs a hyperspectral image camera capable of detecting electromagnetic radiation from 400 nm-1000 nm with a spectral resolution of 204 bands. The CNN algorithm used in this study is based on a modified VGG-16 architecture. The total amount of data from granite samples employed amounts to 19, 456 files. 80% of the total data were used as learning data and the remaining 20% as testing and validation data. The results show that the system has an average precision of over 94%. The study, therefore, concludes that the conjunction of hyperspectral imaging with CNN is a practicable process for the evaluation of the degree of weathering granite without the need for specialised human expertise or prejudice.