Research to automate merging 3D point cloud data of geometries requiring 4π measurement

Reference No.	2024a007
Type/Category	Grant for General Research- Short-term Visiting Researcher
Title of Research Project	Research to automate merging 3D point cloud data of geometries requiring 4π measurement
Principal Investigator	Takenori Sumi（Technology and Incubation Space, Asahi Heat Treatment Co.,ltd. ・None）
Research Period	April 1,2024. - April 3,2024. September 30,2024. - October 2,2024. March 26,2025. - March 31,2025.
Keyword(s) of Research Fields	Point cloud, Registration
Abstract for Research Report	This study aims to automate merging 3D point cloud data acquired by optical 3D measurement devices (3D sensors). Typically, 3D sensors cannot acquire information on the position of the camera's blind spot. Therefore, in order to obtain the entire shape of the object as 3D point cloud data, it is necessary to acquire point cloud data that compensates for the blind spots of the camera from each other and merge them appropriately. In the surveying field, where the utilization of point cloud processing is advancing, the technology for merging complementary point cloud data, including automatic merging, has advanced to the point where it has become a product and service. However, this is not the case for the manufacturing industry. In this case, all of the complementary point cloud data can only be obtained by acquiring information on the shading spots to the camera, and this restriction creates difficulties different from those of automatic merging in the surveying field. For example, in the surveying field, the position can be aligned with reference to the target. However, in the manufacturing industry, as long as the target is grounded for measurement, an operation is required to turn the target over in order to obtain information on the shading region. This operation causes misalignment with the assumed target, making direct assistance difficult. Therefore, restoring the object shape is performed by merging point clouds passed through the registration process while referring to each other's point clouds. In this restoration process, the point clouds must be adequately close to each other. This manual selection introduces non-verbal parameters and does not always guarantee the validity and reproducibility of the results. We aim to overcome this issue. As a result of this research, we expect to obtain a reproducible prescription for restoring 3D point cloud data with the overall shape that should be obtained, for the case where information on the shading region from the camera is required, simply by specifying data and parameters, i.e., avoiding the imperfections, errors, and effort by skilled person inherent in the manual procedure.
Organizing Committee Members (Workshop) Participants (Short-term Joint Usage)	Takenori Sumi（Asahi Heat Treatment Co.,Ltd.・None）