cbbhuxx / GeoCluster

Enhancing Visual Place Recognition in Spatial Domain on Aerial Vehicle Platforms.

GeoCluser

Enhancing Visual Place Recognition in Spatial Domain on Aerial Vehicle Platforms.

Aerial localisation

Prerequisites

• torch
• torchvision
• opencv-python
• tensorboard
• matplotlib
• sklearn
• numpy

Please see "requirements.txt" for a detailed list of packages and versions.

The code expects data to be in the following directory structure:

experiment\
├── features
|   ├── Beijing
|   |   ├── Beijing_feature_GeoCluser_1.00.npy
|   |   ├── Beijing_feature_GeoCluser_0.85.npy
|   |   ├── Beijing_feature_GeoCluser_0.85.npy
├── map
|   ├── Beijing.tif   
├── query_images
|   ├── Beijing_query
|   |   ├── 0.png
|   |   ├── 1.png
├── route
|   ├──Beijing_route.npy
├── tiles
|   ├── Beijing_ref_scale_0.85
|   |    ├── 0.png
|   |    ├── 1.png
|   ├── Beijing_ref_scale_1.00
|   |    ├── 0.png
|   |    ├── 1.png
|   ├── Beijing_ref_scale_1.35
|   |    ├── 0.png
|   |    ├── 1.png

Note: scale_1.00, scale_0.85, scale_1.35 refer to the scale of the map tile image and the camera image.

Quick start

If there is no database, put the database image into experiment/tiles/:

python main.py --model=GeoCluster --mode=Pre

If the database already exists:

python main.py --model=GeoCluster --mode=PF

License + attribution/citation

When using code within this repository, please refer the following paper in your publications:

@ARTICLE{10423811,
  author={Chen, Chao and He, Mengfan and Wang, Jun and Meng, Ziyang},
  journal={IEEE Robotics and Automation Letters}, 
  title={GeoCluster: Enhancing Visual Place Recognition in Spatial Domain on Aerial Vehicle Platforms}, 
  year={2024},
  volume={9},
  number={3},
  pages={3013-3020},
  keywords={Feature extraction;Visualization;Task analysis;Training;Location awareness;Databases;Autonomous aerial vehicles;Vision-based navigation;localization;recognition;deep learning for visual perception},
  doi={10.1109/LRA.2024.3363536}}