Simple-shaped object pose estimation based on the grounding face

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K19830
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 61010:Perceptual information processing-related
• Research Institution: Meijo University
• Principal Investigator: TASAKI Tsuyoshi 名城大学, 理工学部, 准教授 (40824660)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 姿勢推定 / インスタンスセグメンテーション

Outline of Final Research Achievements

There is a demand for robots in retail stores to automate the display of goods. However, most goods in retail store have few shape features. Therefore, it is difficult for conventional methods to estimate poses of the goods. This study proposes a new method to estimate the grounding face of the good. The method simplifies the three-dimensional pose estimation problem into the one-dimensional pose estimation problem to estimate the yaw angle on the grounding face. The ratio of estimating the pose with an angle error of 5 degree is 99.9%, which is 71.1 pt higher than that of a conventional method.

Free Research Field

ロボットビジョン

Academic Significance and Societal Importance of the Research Achievements

物体の姿勢推定を高精度化できたことで、ロボットによる商品整理作業の実現可能性を示し、労働力不足という社会問題を解決に近づけることができた。さらに、三次元的な物体姿勢を、物体の接地面と接地面上の一次元の回転で表現することで、姿勢推定精度が向上するという知見は、他の物体姿勢推定手法にも応用できる。したがって、本研究成果で得られた知見は、物体姿勢推定という学術分野の発展に貢献できると考えられる。