This project focuses on estimating the pose of an Unmanned Ground Vehicle (UGV) from an Unmanned Aerial Vehicle (UAV) using vision-based techniques.
Instead of relying on fiducial markers like AprilTag or ArUco, this approach explores natural landmark detection and feature-based methods, making the system more robust and deployable in real-world environments.
UAV (Camera Platform)
Captures top-down imagery Detects UGV using: Feature matching OR Neural network
UGV (Target)
No markers required Recognized via natural features
Pipeline
Image acquisition Feature extraction / inference Feature matching or detection Pose estimation Output relative pose (position + orientation)
Feature detection & matching Visual odometry Perspective-n-Point (PnP) End-to-end deep learning for localization
- git clone https://github.com/caanges/Slarc_1.git
- cd Slarc_1
- pip install -r requirements.txt
- Christoffer Angestam
- Emil Ekengren
- Elwin Green
- Edvin Mörk
- Malek Saleh