A study on the autonomous detection of impact craters

Aburaed, Nour and Alsaad, Mina and Mansoori, Saeed Al and Al-Ahmad, Hussain; El Gayar, Neamat and Trentin, Edmondo and Ravanelli, Mirco and Abbas, Hazem, eds. (2022) A study on the autonomous detection of impact craters. In: Artificial Neural Networks in Pattern Recognition - 10th IAPR TC3 Workshop, ANNPR 2022, Proceedings. Lecture Notes in Computer Science . Springer Science and Business Media Deutschland GmbH, ARE, pp. 181-194. ISBN 9783031206498 (https://doi.org/10.1007/978-3-031-20650-4_15)

[thumbnail of Aburaed-etal-ANNPR-2022-A-study-on-the-autonomous-detection-of-impact] Text. Filename: Aburaed_etal_ANNPR_2022_A_study_on_the_autonomous_detection_of_impact.pdf
Accepted Author Manuscript
Restricted to Repository staff only until 11 November 2024.
License: Strathprints license 1.0

Download (3MB) | Request a copy

Abstract

Planet surface studies is one of the most popular research areas in planetary science, as it is useful to attain information about a planet's history and geology without directly landing on its surface. Autonomous detection of craters has been of particular interest lately, especially for Mars and Lunar surfaces. This review study deals with the technical implementation, training, and testing of YOLOv5 and YOLOv6 to gauge their efficiency in detecting craters. YOLOv6 is the most recent member of the YOLO family, and it is believed that it outperform all of its predecessors. In addition to comparing the aforementioned two models, the performance of the most widely used optimization functions, including SGD, Adam, and AdamW is studied as well. The methods are evaluated using mAP and mAR to verify whether YOLOv6 potentially outperforms YOLOv5, and whether AdamW is capable to generalize better than its peer optimizers.