Abstract:Rapid ice recession in the Arctic Ocean, with predictions of ice-free summers by 2060, opens new maritime routes but requires reliable navigation solutions. Current approaches rely heavily on subjective expert judgment, underscoring the need for automated, data-driven solutions. This study leverages machine learning to assess ice conditions using ship-borne optical data, introducing a finely annotated dataset of 946 images, and a semi-manual, region-based annotation technique. The proposed video segmentation model, UPerFlow, advances the SegFlow architecture by incorporating a six-channel ResNet encoder, two UPerNet-based segmentation decoders for each image, PWCNet as the optical flow encoder, and cross-connections that integrate bi-directional flow features without loss of latent information. The proposed architecture outperforms baseline image segmentation networks by an average 38\% in occluded regions, demonstrating the robustness of video segmentation in addressing challenging Arctic conditions.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: With the rapid retreat of Arctic sea ice, opening up new sea routes has become possible, but reliable navigation solutions are crucial for ensuring the safety of voyages. Current methods mainly rely on the subjective judgment of experts, which highlights the need for automated, data - driven solutions. This paper aims to evaluate ice conditions by using ship - borne optical data through machine - learning methods to provide more reliable and objective navigation support. Specifically, this research addresses the following key issues: 1. **Limitations of existing methods**: - Current ice - condition assessment relies on the subjective judgment of experts and lacks objectivity and consistency. - Existing image - segmentation methods (such as K - means clustering, Otsu method, etc.) perform poorly in complex scenes, especially when it is difficult to accurately distinguish between ice and non - ice areas under changing lighting conditions, contrast differences and noise impacts. - There is a lack of publicly available high - quality labeled datasets, resulting in insufficient data for training models or rough labeling. 2. **Lens - occlusion problem**: - Images taken by ship - borne cameras are often affected by lens occlusions such as water droplets, resulting in the loss of information in some areas. Existing processing methods (such as spatial interpolation) have limited effects in occluded areas with complex geometric structures. 3. **Utilization of time information**: - Existing methods fail to fully utilize the time information in videos to improve segmentation accuracy, especially in the prediction of occluded areas. To solve these problems, this paper proposes a new method based on video segmentation - UPerFlow. This method combines optical - flow estimation and semantic - segmentation techniques, can better handle occluded areas and improves the overall segmentation performance. In addition, the author has also developed a medium - scale finely - labeled dataset for training and evaluating the model. ### Main contributions 1. **New labeling method**: A medium - scale, finely - labeled dataset is proposed, which contains 946 images and covers six types of objects: floating ice, broken ice, water, ship, sky and iceberg. 2. **Baseline evaluation**: Existing image - semantic - segmentation networks are evaluated and an accurate classifier baseline is established. 3. **Improved video - semantic - segmentation architecture**: The UPerFlow model is introduced. By combining optical - flow information and multi - scale feature fusion, the segmentation performance in occluded areas is significantly improved. These improvements make the model more robust and reliable in navigation applications under complex ice conditions.

Breaking The Ice: Video Segmentation for Close-Range Ice-Covered Waters

Panoptic Segmentation for Seasonal River Scene Based on Spatial Context Prior and DenseASPP

A Multi-Task Network with Dynamic Segmentation for Sea Ice Classification in Arctic Shipping Route Optimization

Real-time Water Area Segmentation for USV using Enhanced U-Net

An Open-Source Platform For Underwater Image And Video Analytics

Segmenting sea ice floes in close-range optical imagery with active contour and foundation models

Fine-Scale Sea Ice Segmentation for High-Resolution Satellite Imagery with Weakly-Supervised CNNs

Enhancing sea ice segmentation in Sentinel-1 images with atrous convolutions

Predicting ice flow using machine learning

RUF: Effective Sea Ice Floe Segmentation Using End-to-End RES-UNET-CRF with Dual Loss

CSEU-Net: ConvNeXt-SE-U-Net for river ice floe segmentation using unmanned aerial vehicle grayscale remote sensing images

Collision-Free Waterway Segmentation for Inland Unmanned Surface Vehicles

Precise and Fast Segmentation of Sea Ice in High-Resolution Images Based on Multiscale and Knowledge Distillation

Towards automated identification of ice features for surface vessels using deep learning

Integrating a data-driven classifier and shape-modulated segmentation for sea-ice floe extraction

Decoding the Partial Pretrained Networks for Sea-Ice Segmentation of 2021 Gaofen Challenge

An image augmentation‐based ice monitoring method for safe navigation of polar ships

Automatic Extraction of Floating Ice at Antarctic Continental Margin from Remotely Sensed Imagery Using Object-Based Segmentation

The Identification of Ice Floes and Calculation of Sea Ice Concentration Based on a Deep Learning Method

Sea Ice Extraction via Remote Sensed Imagery: Algorithms, Datasets, Applications and Challenges

An Automated System to Monitor River Ice Conditions Using Visible Infrared Imaging Radiometer Suite Imagery