ICI-YOLOv8 Rapid Identification of Antarctic Sea Ice Cracks and Numerical Analysis of Monte Carlo Simulation Under Probability Distribution
Year: 2025
Type: article
Abstract: Labeling ice cracks in Antarctic near-shore sea ice aerial orthophotos is critical for sea ice cargo route development; rapid, accurate identification and labeling of cracks in UAV imagery aids safe goods transfer between icebreakers and expedition stations, and studying ice crack distribution provides a key basis for assessing sea ice route reliability. Ice cracks have complex morphologies that traditional recognition methods struggle to handle, so this study proposes the ICI-YOLOv8 algorithm to improve sea ice crack detection near Antarctica’s Zhongshan Station, using crack density and fractal dimension to characterize spatial distribution and a Monte Carlo-based numerical model to quantify distribution probability. The algorithm achieves 0.628 accuracy and 0.662 [email protected] (outperforming comparable methods in speed and accuracy) and reaches 0.933 accuracy and 0.657 [email protected] with better generalization than similar models when tested on general remote sensing water datasets; a positive correlation exists between fractal dimension and ice crack density, and Monte Carlo simulation and probability distribution models verify their distribution properties. The proposed algorithm is suitable for rapid summer Antarctic near-shore sea ice crack identification, the numerical model effectively quantifies crack distribution to aid route development, and this study is important for understanding polar ice stability and sea ice route development.
Source: Remote Sensing
DOI: 10.3390/rs17213646
Authors: Xiaomin Chang, Lulin Zhang, Yuchen Wang, Fuqiang Li, Xu Yao, Yinke Dou
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,258
Awards: G2977255755, G8330465079
OpenAlex ID: W4415924568