@Article{isprs-archives-XLVI-2-W1-2022-73-2022,
AUTHOR = {Bellavia, F. and Morelli, L. and Menna, F. and Remondino, F.},
TITLE = {IMAGE ORIENTATION WITH A HYBRID PIPELINE ROBUST TO ROTATIONS AND WIDE-BASELINES},
JOURNAL = {The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
VOLUME = {XLVI-2/W1-2022},
YEAR = {2022},
PAGES = {73--80},
URL = {https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XLVI-2-W1-2022/73/2022/},
DOI = {10.5194/isprs-archives-XLVI-2-W1-2022-73-2022}
}

To replicate the results obtained in the paper, you have to match the keypoints found with the various feature extractors tested (see the related projects on GitHub), generate the sparse model with COLMAP and run the main.py script contained in the TargetTriangulation repository. Repository for the hand-crafted methods (SIFT, SURF and AKAZE):

Repositories for the learning-based methods: https://github.com/vcg-uvic/lf-net-release https://github.com/rpautrat/SuperPoint https://github.com/naver/r2d2 https://github.com/axelBarroso/Key.Net https://github.com/lzx551402/ASLFeat

Others interesting learning-based keypoint extractors not tested in the paper: https://github.com/kimphys/UnsuperPoint.pytorch https://github.com/TRI-ML/KP2D https://github.com/ignacio-rocco/sparse-ncnet

Many other interesting projects related to image matching: https://github.com/shamangary/awesome-local-global-descriptor

This folder contains a script useful to triangulate targets marked on images starting from a COLMAP sparse cloud. Edit the config.py file to set the input/output paths, then run the "main.py" script. NOTE: This script manage only numerical target ID.

Open Anaconda prompt as administrator

>conda create --name opencv
>conda activate opencv
>conda install -c conda-forge opencv
>conda install -c anaconda pandas
>conda install -c conda-forge matplotlib

>conda activate opencv
>python main.py