Computer vision enhances the potential of ultrabroadband imaging in non-destructive testing

    Led by Assistant Prof. Kou Li, a research group in Chuo University (President: KAWAI Hisashi, Hachioji-shi, Tokyo, Japan), under a collaboration with National Institute of Informatics (Director-General: KUROHASHI Sadao, Chiyoda-ku, Tokyo, Japan), has developed a synergetic strategy among millimeter-wave–terahertz–infrared photo-monitoring and computer vision three-dimensional modelling toward ubiquitous non-destructive inspections, with a recent paper publication in Advanced Materials Technologies.

    While non-destructive inspections of daily necessities and industrial products require material composition identifications and structural reconstructions of composite multi-layered three-dimensional (3D) objects, the development of analyzing systems that fully satisfy those performances is still insufficient. Particularly, synergetic combinations are urgently indispensable among non-destructive permeable inner material identifications with millimeter-wave (MMW)–terahertz (THz)–infrared (IR) photo-monitoring and computer-vision (CV)-driven 3D modelling. However, respective approaches on the above individual concept crucially regulate their use as planar image acquisition operations for the former (missing cross-sectional information) and visual light (Vis)-based outer views for the latter (lacking non-destructive permeability to inner structures).

    To this end, this manuscript made the following significant contributions.

1. Integrating multi-wavelength MMW–IR monitoring functions into carbon nanotube (CNT) film photo-thermoelectric (PTE) imagers, which works in ultrabroadband ranges over conventional wide-band sensors at comparable sensitivities with those of existing narrow-band detectors in each region.
    
2. Hybrid multi-CV operations among visual hull (VH) and Computed tomography (CT) with the CNT film PTE imager-based MMW–IR monitoring system as follows: speculating 3D spatial locations of each constituent material of targets by VH (shadow/silhouette-based vague views) as a first step of non-destructive inspections, then restores their detailed cross-sectional structures (e.g., columnar? or hollow?) by CT.
    
3. Demonstrating material-identifying non-destructive structural reconstructions against multi-layered composite opaque 3D target objects by incorporating the aforementioned fundamental device and system preparations.
    
4. Satisfying fundamental performances (1 mm resolution at 150 sec. for VH and 2 min. for CT) comparable to existing Vis-reconstruction methods by the presenting system, while the CNT film PTE imager fully functions over 1.15 mm–660 nm wavelength ranges beyond the above conventional CV approaches.

    The paper was published online in the international scientific journal, Advanced Materials Technologies (February 15, 2025).

<Researcher>

Kou Li, Assistant Professor, Department of Electrical, Electronic, and Communication Engineering of the Faculty of Science and Engineering, Chuo University　

Daiki Shikichi, Master course student, Department of Electrical, Electronic, and Communication Engineering, Graduate School of Science and Engineering, Chuo University

Miki Kubota, Master course student, Department of Electrical, Electronic, and Communication Engineering, Graduate School of Science and Engineering, Chuo University

Yukio Kawano, Professor, Department of Electrical, Electronic, and Communication Engineering of the Faculty of Science and Engineering, Chuo University

Raito Ota, Bachelor student (at the time of research), Department of Electrical, Electronic, and Communication Engineering of the Faculty of Science and Engineering, Chuo University

Takeru Q. Suyama, Ph.D. student, National Institute of Informatics, SOKENDAI

Satoshi Ikehata, Associate Professor, Digital Content and Media Sciences Research Division, National Institute of Informatics

Imari Sato, Professor, Digital Content and Media Sciences Research Division, National Institute of Informatics

<Authors> 

^†Co-first author, ^*Corresponding author

Daiki Shikichi^†,1, Raito Ota^†,1, Miki Kubota^†,1, Yuya Kinoshita¹, Noa Izumi¹, Mitsuki Kosaka¹, Tomoki Nishi¹, Daiki Sakai¹, Yuto Matsuzaki¹, Leo Takai¹, Minami Yamamoto¹, Yuto Aoshima¹, Ryoga Odawara¹, Takeru Q. Suyama^2,3, Hiroki Okawa⁴, Zhenyu Zhou³, Tomoya Furukawa^5,6, Shota Wada^5,6, Satoshi Ikehata³, Imari Sato³, Yukio Kawano*^,1,3,4, and Kou Li*^,1

<Affiliation>

¹Department of Electrical, Electronic, and Communication Engineering, Faculty of Science and Engineering, Chuo University

²The Graduate University for Advanced Studies, SOKENDAI

³National Institute of Informatics

⁴Kanagawa Industrial Institute of Science and Technology

⁵Laboratory for Future Interdisciplinary Research of Science and Technology, Tokyo Institute of Technology

⁶Department of Electrical and Electronic Engineering, School of Engineering, Tokyo Institute of Technology