Year: 2025
Type: article
Abstract: This paper compares and analyzes the performance of deep learning-based object detection models for mine detection, particularly Faster R-CNN and YOLOv5, using side-scan sonar images. Additionally, it proposes effective data augmentation method to enhance the generalization performance of mine detection models. Performance evaluation based on the structure and size of each model indicates that the two-stage model, Faster R-CNN, is more suitable for precise search tasks, while the one-stage model, YOLOv5, offers faster processing speed, making it advantageous for rapid mine detection in large maritime areas. This contributes significantly to improving the efficiency of maritime boundary missions and operations utilizing autonomous underwater vehicles, thereby making a substantial impact on naval operations and mine counter measures strategies.
Authors: Jonghyeon Mun, Sieon Park, Jaehwan Kim, Changhyung Kim, Daeyeol Kim, Chaebong Sohn
Institutions: Kwangwoon University, Kyungnam University
Language: English
Subfield: Oceanography
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4409202685