CRESENT: a CRatEr-baSed pose estimation datasEt for cisluNarlocated spacecrafT

CRESENT is first introduced in the paper “Robust Perspective-n-Crater for Crater-based Camera Pose Estimation” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops.

Overview:

This dataset contains images produced by The University of Adelaide using PANGU Planet Surface Simulation Software developed by the Space Technology Centre at the University of Dundee, Scotland.

High-resolution lunar DEMs from the PDS data node were rendered in PANGU, and images were taken above the lunar surface at an altitude of 100km with varying angles off nadir to mimic the expected conditions of a lunar orbiter surface surveillance mission.

The dataset contains images taken above four different surface regions on the Moon, each within a region of 45 degrees latitude and 45 degrees longitude.

Data organisation:

There are four root folders, each containing images produced under one of the four lunar regions:

LDEM_x_yE_l_mN/

where x and y are the latitude bounds (degrees) of the lunar region and l and m are the longitude bounds (degrees) of the lunar region, rendered in PANGU. Note that due to the high resolution of the DEMs, any region of the Moon that was outside these latitude and longitude bounds was not rendered (and the surface will appear cut off/black at the boundaries of these regions).

Within each of the lunar region folders, there are seven subfolders:

LDEM_x_yE_l_mN_float_60fov_1024_1024_ideg_off_nadir/

where i is the viewing angle (degree) off nadir the image was taken at, where i is either 0, 10, 20, 30, 40, 50, or 60 degrees. Each subfolder contains a folder of images, a poses.csv file of ground truth poses, a calibration file and a file detailing the specifics of the rendered LDEM. Note that each image taken within each sub-directory of each lunar region folder will have the same number of files, each located at the same position in the Selenographic reference frame, but at different angles off nadir.  For example, LDEM_-90_-45E_0_45N/LDEM_-90_-45E_0_45N_float_60fov_1024_1024_0deg_off_nadir/ and  LDEM_-90_-45E_0_45N/LDEM_-90_-45E_0_45N_float_60fov_1024_1024_60deg_off_nadir/ will both have the same number of images in their images/ subdirectory, and each image file number in these directories was taken at the same camera position, but at viewing angles of 0 degrees and 50 degrees off nadir, respectively.

Each line in the poses file contains the X, Y, Z position (m) and the yaw, pitch, roll angles (degrees) of the camera in the selenographic reference frame. There are the same number of lines in the poses file as images in each subfolder, e.g., the first line of the poses file corresponds to the XYZ yaw pitch roll pose of the camera that generated images/0.png, the second line of the poses file corresponds to the pose of the camera of images/1.png, etc.

How to cite:

Users of this dataset are requested to cite the following paper.

Reference String

McLeod, S. et al. Robust Perspective-n-Crater for Crater-based Camera Pose Estimation in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops (June 2024)

Bibtex

@InProceedings{Mcleod_2024_CVPR,

author = {Mcleod, Sofia and Chng, Chee Kheng and Ono, Tatsuharu and Shimizu, Yuta and Hemmi, Ryodo and Holden, Lachlan and Rodda, Matthew and Dayoub, Feras and Miyamoto, Hirdy and Takahashi, Yukihiro and Kasai, Yasuko and Chin, Tat-Jun},

title = {Robust Perspective-n-Crater for Crater-based Camera Pose Estimation},

booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},

month = {June},

year = {2024},

pages = {6760-6769} 

}

DISCLAIMER:

The dataset provided is a lower resolution version of the original and is intended for academic research purposes only, consistent with the terms of the Open Access License that applies to the usage of this dataset. Please contact tat-jun.chin@adelaide.edu.au if you require the higher resolution versions of the dataset.