A Parallel Workflow for Polar Sea-Ice Classification using Auto-labeling of Sentinel-2 Imagery

Jurdana Masuma Iqrah,Wei Wang,Hongjie Xie,Sushil Prasad

2024-03-20

Abstract:The observation of the advancing and retreating pattern of polar sea ice cover stands as a vital indicator of global warming. This research aims to develop a robust, effective, and scalable system for classifying polar sea ice as thick/snow-covered, young/thin, or open water using Sentinel-2 (S2) images. Since the S2 satellite is actively capturing high-resolution imagery over the earth's surface, there are lots of images that need to be classified. One major obstacle is the absence of labeled S2 training data (images) to act as the ground truth. We demonstrate a scalable and accurate method for segmenting and automatically labeling S2 images using carefully determined color thresholds. We employ a parallel workflow using PySpark to scale and achieve 9-fold data loading and 16-fold map-reduce speedup on auto-labeling S2 images based on thin cloud and shadow-filtered color-based segmentation to generate label data. The auto-labeled data generated from this process are then employed to train a U-Net machine learning model, resulting in good classification accuracy. As training the U-Net classification model is computationally heavy and time-consuming, we distribute the U-Net model training to scale it over 8 GPUs using the Horovod framework over a DGX cluster with a 7.21x speedup without affecting the accuracy of the model. Using the Antarctic's Ross Sea region as an example, the U-Net model trained on auto-labeled data achieves a classification accuracy of 98.97% for auto-labeled training datasets when the thin clouds and shadows from the S2 images are filtered out.

Computer Vision and Pattern Recognition,Distributed, Parallel, and Cluster Computing,Machine Learning

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the classification of polar sea ice using Sentinel - 2 satellite images. Specifically, the research aims to develop a powerful, effective, and scalable system for classifying polar sea ice as thick - snow - covered, young - thin, or open water. Since the Sentinel - 2 satellite is actively capturing high - resolution images of the Earth's surface, there are a large number of images that need to be classified. One of the main obstacles is the lack of labeled Sentinel - 2 training data (images) as ground truth labels. To solve this problem, the research proposes an automatic labeling method based on color thresholds and achieves a 9 - fold data - loading acceleration and a 16 - fold map - reduction acceleration using PySpark through a parallel workflow. In addition, to improve the efficiency of model training, the research also adopts the Horovod framework to distribute the training of the U - Net model on 8 GPUs, achieving a 7.21 - fold speedup without affecting the accuracy of the model. Finally, using the Ross Sea region in Antarctica as an example, the U - Net model trained on automatically - labeled data achieved a 98.97% classification accuracy on Sentinel - 2 images after thin clouds and shadows were removed.

A Parallel Workflow for Polar Sea-Ice Classification using Auto-labeling of Sentinel-2 Imagery

Toward Polar Sea-Ice Classification using Color-based Segmentation and Auto-labeling of Sentinel-2 Imagery to Train an Efficient Deep Learning Model

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

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

Data Augmentation for SAR Sea Ice and Water Classification Based on Per-Class Backscatter Variation With Incidence Angle

Geo-knowledge-informed Deep Learning for Auto-identification of Supraglacial Lakes on the Greenland Ice Sheet from Satellite Imagery.

Partial Label Learning with Focal Loss for Sea Ice Classification Based on Ice Charts

A Multi-scale Dual Attention Network for the Automatic Classification of Polar Sea Ice and Open Water Based on Sentinel-1 SAR Images

DF-UHRNet: A Modified CNN-Based Deep Learning Method for Automatic Sea Ice Classification from Sentinel-1A/B SAR Images

Region-level labels in ice charts can produce pixel-level segmentation for Sea Ice types

ArcticNet: A Deep Learning Solution to Classify Arctic Wetlands

A Self-Trained Model for Cloud, Shadow and Snow Detection in Sentinel-2 Images of Snow- and Ice-Covered Regions

A Novel Method for Automated Supraglacial Lake Mapping in Antarctica Using Sentinel-1 SAR Imagery and Deep Learning

Automated Extraction of Antarctic Glacier and Ice Shelf Fronts from Sentinel-1 Imagery Using Deep Learning

Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks

Automated surface melt detection over the Antarctic from Sentinel-1 imagery using deep learning

Detection of Antarctic Surface Meltwater Using Sentinel-2 Remote Sensing Images via U-Net With Attention Blocks: A Case Study Over the Amery Ice Shelf

IceGCN: An Interactive Sea Ice Classification Pipeline for SAR Imagery Based on Graph Convolutional Network

Classification of sea ice types in Sentinel-1 synthetic aperture radar images

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

Deep learning based F-Mask alternative for Sentinel-2 images in polar regions