Triangulate geometric constraint combined with visual-flow fusion network for accurate 6DoF pose estimation

作者：

Highlights：

• The TGCPose6D method is composed of keypoint detection and translation refinement stages.

• The accuracy of keypoint location is improved by Triangulate Geometric Constraint vectors.

• The Visual-Flow Fusion network is designed to refine the prediction of translation.

• The pose prediction accuracy of TGCPose6D is improved at least 11.8% by TGC vectors.

• The accuracy of pose has increased by 9.58% through VFFNet without depth supervision.

摘要

•The TGCPose6D method is composed of keypoint detection and translation refinement stages.•The accuracy of keypoint location is improved by Triangulate Geometric Constraint vectors.•The Visual-Flow Fusion network is designed to refine the prediction of translation.•The pose prediction accuracy of TGCPose6D is improved at least 11.8% by TGC vectors.•The accuracy of pose has increased by 9.58% through VFFNet without depth supervision.

论文关键词：6D object pose estimation,Iterative translation refinement,Triangulate geometric constraint,Visual-flow feature fusion

论文评审过程：Received 22 September 2020, Revised 2 February 2021, Accepted 3 February 2021, Available online 17 February 2021, Version of Record 25 February 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2021.104127