Related Experiment Video
Updated: Jun 29, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
530
Spacecraft Homography Pose Estimation with Single-Stage Deep Convolutional Neural Network.
Shengpeng Chen1, Wenyi Yang1, Wei Wang1
1School of Aeronautics and Astronautics, Sun Yat-sen University, Shenzhen 510275, China.
Sensors (Basel, Switzerland)
|March 28, 2024
Summary
This study introduces a novel single-stage deep convolutional neural network for spacecraft pose estimation. The method simplifies complex operations, enabling efficient and accurate pose determination in space environments.
Area of Science:
- Computer Vision
- Robotics
- Automation Systems
Background:
- Spacecraft pose estimation is crucial for autonomous operations in space.
- Existing multi-stage methods are complex and computationally intensive, posing challenges in extreme space environments.
Purpose of the Study:
- To propose a novel, efficient single-stage deep convolutional neural network for spacecraft homography pose estimation.
- To address the limitations of existing complex, multi-stage pose estimation techniques.
Main Methods:
- Formulation of a homomorphic geometric constraint equation for spacecraft with planar features.
- Employment of a single-stage 2D keypoint regression network for homography 2D keypoint coordinate extraction.
- Decomposition of the homography matrix for rough pose estimation, followed by pixel error-based loss function for pose refinement.
Main Results:
- Demonstrated effectiveness of the proposed single-stage network for spacecraft pose estimation.
- Achieved competitive or superior performance compared to state-of-the-art methods on widely used datasets.
Conclusions:
- The proposed single-stage deep convolutional neural network offers an effective and simplified approach to spacecraft pose estimation.
- This method advances automation systems, control theory, and robot technology for space applications.
