摘要
为减少由冬季寒潮导致的海冰灾害损失,利用遥感手段对我国高纬度海域海冰风险进行监测,并提供决策支撑。文章利用环境减灾二号A/B卫星的16 m分辨率多光谱数据,采用基于DeepLab V3模型的语义分割方法,对渤海辽东湾地区的海冰范围进行提取和动态分析,得出海冰范围的动态变化规律,并进行提取精度评估。结果表明:利用该方法的海冰提取准确率可以达到97.4%以上,可以满足海冰的日常监测需要,验证了环境减灾二号A/B卫星在风险要素遥感监测的应用能力。
The cold wave in winter can cause sea ice disaster,remote sensing is used to monitor the sea ice risk in China’s high latitude sea areas and provide decision support.Using the 16 meter resolution multispectral data of HJ-2A/B satellites and the semantic segmentation method based on Deeplab V3 model,this paper extracts and dynamically analyzes the sea ice range in Liaodong Bay,Bohai Sea,obtains the dynamic change law of sea ice range,and evaluates the extraction accuracy.The results show that the accuracy of sea ice extraction using this method can reach more than 97.4%,which can meet the needs of daily sea ice monitoring,and verify the application ability of HJ-2A/B satellites in remote sensing monitoring of risk factors.
作者
陈璐
胡凯龙
刘明博
CHEN Lu;HU Kailong;LIU Mingbo(National Disaster Reduction Center of China, MEM, Beijing 100124, China)
出处
《航天器工程》
CSCD
北大核心
2022年第3期147-152,共6页
Spacecraft Engineering
基金
国家重点研发计划(2021YFB3901205)。
关键词
环境减灾二号卫星
海冰监测
深度学习
语义分割
HJ-2A/B satellites
sea ice monitoring
deep learning
semantic segmentation
