Holorecognizer: implementation expansion of machine learning to optical setup for three-dimensional recognition

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K12012
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61010:Perceptual information processing-related
• Research Institution: Kitami Institute of Technology
• Principal Investigator: Sugisaka Jun-ichiro 北見工業大学, 工学部, 准教授 (00599227)
• Co-Investigator(Kenkyū-buntansha): 安井 崇 北見工業大学, 工学部, 准教授 (20403438)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 人工知能 / 機械学習 / ホログラム / 逆散乱問題 / 計算電磁気学

Outline of Final Research Achievements

We have developed a novel system that can discriminate the three-dimensional structure of an object by using the diffraction of lightwave. The convex or concave defect on a sample surface is discriminated by transmitting the scattered light from the sample through the optical filter. The optical filter can accurately extract the three-dimensional information of the sample from the scattered light. We also designed a novel artificial-intelligence system that consists of an optical system and electronic computer, including a machine learning algorithm that simultaneously trains (designs) the optical filter and the electronic computer to obtain optimal discrimination results.

Free Research Field

光学

Academic Significance and Societal Importance of the Research Achievements

物体表面の微細な凹凸の判別は、半導体ウェハの欠陥検査など、従来の技術では困難であった計測手法の開発が期待できる。本研究で開発した光フィルタは、散乱光がフィルタを透過する過程で機械学習と同等の演算を実現でき、通常は電子計算機のプログラムとして実装される機械学習を、新たな形態で実現するものである。このシステムの利点は、必要な演算処理をきわめて高速・低消費電力で実現でき、エネルギー効率の点でも優れた人工知能の研究開発にもつながるものである。