This is the offical repository of the paper:
Cross-view SLAM solver: global pose estimation of monocular ground-level video frames for 3D reconstruction using a reference 3D model from satellite images
@article{elhashash2022,
author = {Mostafa Elhashash and Rongjun Qin},
title = {Cross-view {SLAM} solver: {Global} pose estimation of monocular ground-level video frames for {3D} reconstruction using a reference {3D} model from satellite images},
journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
volume = {188},
pages = {62-74},
year = {2022},
issn = {0924-2716},
doi = {https://doi.org/10.1016/j.isprsjprs.2022.03.018},
}
Execute the following command after modifying the config.yaml
.
Cross_View_SLAM.exe PATH_TO_YAML_FILE
config.yaml
contains all the parameters needed for the program to run. Descriptions for each parameter are provided as comments in the file.
The results folder should contain trajectory.txt
for the estimated poses defined in TUM trjectory format with replacing the timestamp by the image name.
The results also contain the undistorted images and reconstruction results in VisualSFM NVM format. You can use OpenMVS to densify the results. Note we split the reconstruction into a few segments before densification to avoid memory issues.
As mentioned in the paper, the weights alpha
and beta
in Eq. (4) need to be tuned to get good resuls. Refer to the conclusion section for the limitations of this work.
Code release might be possible but will take a longer time.