Abstract:6-D pose estimation is an important branch in the field of vision measurement and is widely used in the fields of robotics, autonomous driving, and reality augmentation. The latest research trend in 6-D pose estimation is to train a deep neural network to directly predict the 2-D projection position of the 3-D keypoint from the image, establish the corresponding relationship, and, finally, use the perspective-n-point (PnP) algorithm to perform pose estimation. The current challenge of pose estimation is that, when objects are textureless, occluded, or scene-cluttered, the detection accuracy is reduced, and most of the existing algorithm models are large and cannot accommodate real-time requirements. In this article, we introduce a densely connected feature pyramid network (DFPN) that can efficiently integrate and utilize features. We combine the cross-stage partial network (CSPNet) with DFPN to design a new network for 6-D pose estimation, DFPN-6-D, a new approach for 6-D object pose estimation. DFPN-6-D can efficiently and accurately handle objects with textureless, occluded, and scene clutter and estimate their full 6-D poses in a single shot. Furthermore, we propose a new confidence calculation method and loss function for object pose estimation, which can fully consider spatial information. Finally, we propose a novel augmentation method for direct 6-D pose estimation approaches to improve performance and generalization ability in the case of occlusion, which is called 6-D augmentation. Our approach achieves a new state-of-the-art accuracy of 98.06 and 87.09 in terms of the ADD(-S) metric on the Linemod dataset and the Occluded-Linemod dataset, and our method also achieves the best result in terms of the different metric on the MULT-I dataset, the BIN-P dataset, and the T-LESS dataset, respectively, while still running end-to-end at over 65 frames/s. The experimental results demonstrate that our algorithm is robust to textureless materials and occlusion while running more efficiently than other methods. We also deploy our proposed method to a real robot to grasp and manipulate objects based on the estimated pose.

6-DOF Pose Estimation for Spacecraft Based on Soft Classification of Euler Angles Optimization

Pose Estimation and Neural Implicit Reconstruction Towards Non-Cooperative Spacecraft Without Offline Prior Information

3D Point-to-Keypoint Voting Network for 6D Pose Estimation

Pose Estimation for Non-cooperative Spacecraft Based on Deep Learning

Monocular 6-Dof Pose Estimation for Non-cooperative Spacecrafts Using Riemannian Regression Network

Pose Estimation for Cross-Domain Non-Cooperative Spacecraft Based on Spatial-Aware Keypoints Regression

Feature-aided Pose Estimation Approach Based on Variational Auto-Encoder Structure for Spacecrafts

Spacecraft Homography Pose Estimation with Single-Stage Deep Convolutional Neural Network

An image appearance based optimization scheme for monocular 6D pose estimation of SOR cabins

Learning-Based Pose Estimation of Non-Cooperative Spacecrafts with Uncertainty Prediction

Pose estimation of non-cooperative spacecraft based on Convelutional Neural Network

Pose Estimation of Large Non-Cooperative Spacecraft Based on Extended PnP Model

A Pose Estimation Method of Space Non-Cooperative Target Based on ORBFPFH SLAM

6D Object Pose Estimation from Approximate 3D Models for Orbital Robotics

Non-Model-Based Monocular Pose Estimation Network for Uncooperative Spacecraft Using Convolutional Neural Network

6D pose estimation of 3D objects in scenes with mutual similarities and occlusions

Pose Determination Of Non-Cooperative Spacecraft Based On Multi-Feature Information Fusion

Uncooperative Spacecraft Pose Estimation Based on Intensity and Range Images Fusion

Real-Time and Efficient 6-D Pose Estimation from a Single RGB Image

DON6D: a decoupled one-stage network for 6D pose estimation

Pose Estimation of Space Objects Based on Hybrid Feature Matching of Contour Points.