Jiahao Guo, Daqi Zhu, and Simon X. Yang
YOLOv10-LBF, sea ice detection, computer vision, lightweight model
Sea ice detection has been regarded as a crucial task in polar environmental monitoring, playing a significant role in climate change research, maritime safety, and resource development. This study proposed YOLOv10-LBF(LightConv-BiFPN-FasterNet), an improved YOLOv10n algorithm, for sea ice detection. The model incorporated a lightweight convolutional structure (LightConv), a bidirectional feature pyramid network (BiFPN), and an efficient neural network module (FasterNet). These enhancements were designed to optimise feature extraction and fusion in the backbone and detection head of the model. Experimental results indicated that fast and accurate sea ice detection in complex backgrounds was achieved. Compared to traditional methods and other YOLO series models, detection precision was improved by 5.3%, and model parameters were reduced by 0.13 MB, demonstrating excellent lightweight performance and adaptability. The YOLOv10-LBF algorithm is expected to provide robust technical support for real-time monitoring and early warning systems, contributing significantly to polar scientific research and environmental conservation.
Important Links:
Go Back
