基于多源数据的北极东北航道冰情示警方法  

Ice warning method of Northeast Passage of Arctic based on multi-source data

作　　者：王翼飞 谢宗轩 WANG Yifei;XIE Zongxuan(Merchant Marine College,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海海事大学商船学院,上海201306

出　　处：《上海海事大学学报》2023年第2期30-37,共8页Journal of Shanghai Maritime University

摘　　要：为减少船舶在北极东北航道航行时发生事故的概率,基于德国不莱梅大学发布的海冰密集度数据和美国国家冰雪数据中心发布的海冰厚度、海冰类型数据建立海冰风险预警模型,对不同空间分辨率海冰数据进行空间投影转换,将北极海域划分为网格并与海冰特征进行对应。将船舶与海冰的位置关系作为聚类特征向量,使用高斯混合模型(Gaussian mixture model,GMM)对网格进行聚类,引入高斯分布重叠率作为评价网格可分性的指标。根据类间可分性较大、类内相似性较高的原则,将海冰网格分成3类,并与其他聚类方法进行对比。实验结果表明,GMM可以很好地根据网格特征差异划分出高风险海冰网格,相比邻域网格聚类方法,其类间可分性更好,精度和稳定性也更好。To reduce the probability of accidents while ships sail in Northeast Passage of Arctic,a sea ice risk warning model is established based on the sea ice concentration data from the University of Bremen in Germany and the sea ice thickness and type data from the National Snow and Ice Data Center of USA.Spatial projection transformation of sea ice data with different spatial resolutions is carried out,the Arctic sea area is divided into grids,and the grids correspond to sea ice characteristics.The position relationship between sea ice and ships is used as the clustering characteristic vector,the Gaussian mixture model(GMM)is used to cluster the grids,and the Gaussian distribution overlap rate is introduced as the index to evaluate the grid separability.Acorrding to the principle of the higher separability between classes and the higher similarity within a class,the sea ice grids are divided into three classes,which is compared with other clustering methods.The experimental results show that GMM can well divide the high-risk sea ice grids according to the difference of grid characteristics,and it has better separability between classes and better accuracy and stability compared with the neighborhood grid cluster method.

关 键 词：海冰密集度 海冰厚度 高斯混合模型(GMM) 冰情示警 

分 类 号：U675.79[交通运输工程—船舶及航道工程]