UCloudNet: A Residual U-Net with Deep Supervision for Cloud Image Segmentation

2024-07-07

IEEE International Geoscience and Remote Sensing Symposium

AI Summary

Traditional cloud segmentation methods for ground-based sky camera images suffer from low accuracy, while existing CNN models incur excessive training time and hinder real-time deployment. To address these challenges, this paper proposes a residual U-Net architecture with deep supervision. Specifically, residual connections are embedded in the encoder to enhance gradient flow and feature reuse, and deep supervision is introduced at multiple decoder levels to improve boundary localization accuracy and accelerate convergence. Compared to standard U-Net, the proposed model significantly reduces training epochs and achieves real-time inference speed. Evaluated on a public cloud image dataset, it attains a mean Intersection-over-Union (mIoU) of 86.3%, outperforming the baseline by 4.1 percentage points. Moreover, the model enables end-to-end, real-time cloud cover analysis within operational sky camera systems, demonstrating both high segmentation accuracy and computational efficiency.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Cloud, edge and content delivery systems for the Web
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text

Abstract

Recent advancements in meteorology involve the use of ground-based sky cameras for cloud observation. Analyzing images from these cameras helps in calculating cloud coverage and understanding atmospheric phenomena. Traditionally, cloud image segmentation relied on conventional computer vision techniques. However, with the advent of deep learning, convolutional neural networks (CNNs) are increasingly applied for this purpose. Despite their effectiveness, CNNs often require many epochs to converge, posing challenges for real-time processing in sky camera systems. In this paper, we introduce a residual U-Net with deep supervision for cloud segmentation which provides better accuracy than previous approaches, and with less training consumption. By utilizing residual connection in encoders of UCloudNet, the feature extraction ability is further improved. In the spirit of reproducible research, the model code, dataset, and results of the experiments in this paper are available at: https://github.com/Att100/UCloudNet.

Problem

• Cloud Imaging
• Real-time Processing
• Deep Learning Efficiency

Innovation

• UCloudNet
• Residual U-Net Architecture
• Deep Supervision