Abstract:
In order to meet the industrial high-precision assembly requirement for components in 3C products, which is texture-less and highly symmetrical, a non-contact pose estima...Show MoreMetadata
Abstract:
In order to meet the industrial high-precision assembly requirement for components in 3C products, which is texture-less and highly symmetrical, a non-contact pose estimation scheme based on point cloud registration is proposed. The scheme use global registration to quickly find the initial coarse pose of the 3C assembly relative to the camera. Considering the singularity problem caused by the high symmetry of the components in 3C products, template-based algorithm is introduced to eliminate the singular pose, then local registration is used to improve the accuracy of the initial pose based on the global registration result. In order to avoid the local registration algorithm falling into local optimum due to insufficient accuracy of the initial pose, we use improved Globally Optimal Iterative Closest Point (Go-ICP) to performs a finer search in the small space including the global optimal pose. The reliability and effectiveness of the scheme are verified by simulation and experiments.
Date of Conference: 06-08 December 2019
Date Added to IEEE Xplore: 20 January 2020
ISBN Information: