| 基于Mask R-CNN 的海洋锋自动识别方法 |
| Automatic Recognition Method of Ocean Front Based on Mask R-CNN |
| |
| DOI：doi：10.3969/j.issn.1003-2029.2025.04.007 |
| 中文关键词: Mask R-CNN 海洋锋 深度学习 目标识别 |
| 英文关键词:Mask R-CNN ocean front deep learning target recognition |
| 基金项目: |
|
| 摘要点击次数: 375 |
| 全文下载次数: 197 |
| 中文摘要: |
| 海洋锋作为大气与海洋相互作用的重要界面，其精确识别对提升气象预报精度、优化海洋
生态系统管理及深化全球气候变化研究具有重要意义。传统基于梯度阈值的海洋锋检测方法因阈值
选择与判别标准的不一致，导致识别精度较低。为此，本文提出了一种基于掩模区域的卷积神经网
络（Mask Region-based Convolutional Neural Network，Mask R-CNN） 的海洋锋自动识别方法，利用
1993—2020 年长时间序列的海表温度数据，实现海洋锋的自动检测与特征提取。对比分析结果表
明：相较于传统梯度阈值法，该方法的整体检测精度平均超过90%，在海洋锋的宽度与强度提取上
误差更小，可识别出更多海洋锋特征，对小尺度特征的识别效果更加显著。本文研究成果有助于深
化对气候与天气变化的理解，从而提升对极端天气事件及全球气候变化的响应能力。 |
| 英文摘要: |
| As an important interface for the interaction between the atmosphere and the ocean, the precise identification of ocean fronts is
of great significance for improving meteorological forecasting accuracy, optimizing marine ecosystem management, and deepening global
climate change research. The traditional ocean front detection method based on gradient threshold has low recognition accuracy due to the
inconsistency between threshold selection and discrimination criteria. To this end, this paper proposes a Mask Region based Convolutional
Neural Network (Mask R-CNN) method for automatic recognition of ocean fronts, which utilizes long-term sea surface temperature data from
1993 to 2020 to achieve automatic detection and feature extraction of ocean fronts. The comparative analysis results show that compared
with the traditional gradient thresholding method, the overall detection accuracy of this method averages over 90%, with smaller errors in
extracting the width and intensity of ocean fronts. It can identify more ocean front features and has a more significant recognition effect on
small-scale features. The results of this paper contribute to a deeper understanding of climate and weather change, thereby enhancing the
ability to respond to extreme weather events and global climate change. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
