Related Experiment Video
Updated: Dec 27, 2025
06:32
Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring
Published on: July 14, 2023
1.7K
3D Pose Estimation for Object Detection in Remote Sensing Images
1State Key Laboratory of Information Engineering in Surveying, Mapping, and Remote Sensing, Wuhan University, Wuhan 430079, China.
Sensors (Basel, Switzerland)
|February 29, 2020
Summary
We introduce Anchor Points Prediction (APP), a novel algorithm for 3D pose estimation in remote sensing images. APP enhances accuracy and efficiency in determining object position and attitude.
Area of Science:
- Computer Vision
- Remote Sensing
- Artificial Intelligence
Background:
- 3D pose estimation in remote sensing images is challenging.
- Existing methods like RoI Transform have limitations in direction information.
Purpose of the Study:
- To develop a new algorithm for accurate 3D object pose estimation in remote sensing.
- To improve upon existing methods for predicting object direction and attitude.
Main Methods:
- Anchor Points Prediction (APP) algorithm predicts multiple feature points using a neural network.
- Establishes homograph transformation between object and image coordinates.
- Introduces a redefined IoU_APP for calculating object direction and posture.
Main Results:
- Achieved accuracy rates of 0.863 on the HRSC2016 dataset and 0.701 on the DOTA dataset.
- Demonstrated significant accuracy improvement compared to previous methods.
- APP algorithm enables efficient one-stage prediction.
Conclusions:
- The APP algorithm provides accurate 3D pose estimation for objects in remote sensing images.
- APP offers a more efficient and easier calculation process due to one-stage prediction.
- This method enhances the ability to determine object position and attitude accurately.
