关键词: 凝聚层次聚类算法, 海冰外缘线, MODIS, 冰水识别

Abstract: In this paper we suggested the importance of sea ice edge in our country's sea ice monitoring and forcasting. The sea ice edge is the boundary between water and ice. Monitoring of sea ice edge's position and the change is vital to sea water exchange, marine fishery, nearshore marine operation, offshore oil-gas exploration, meteorological observation and so on. The conventional method of sea ice edge extraction often has the shortcomings of massive mixed spots existing and the slow computation speed. Clustering is a way to group objects into more than one cluster, so that the same cluster objects have a higher similarity, and an object in a different cluster is larger. Agglomerative hierarchical clustering algorithm is a bottom-up policy. It first considers each object as a cluster, and then combines these clusters into larger cluster, until all the objects in a cluster, or one condition is met. Based on the agglomerative hierarchical clustering algorithm, combined with MODIS images and the characteristics of Bohai suspended sediment distribution, we discussed the MODIS image segmentation algorithm. The algorithm regards spectral characteristics and shape properties as the rule, joining image meshing, breaking grouper and noise removal and other functions, and optimizing the ice-water identification parameters. This algorithm can effectively improve the availability and accuracy of image segmentation results. We combined this algorithm and edge detection algorithms to extract sea ice edge of the image segmentation results. The algorithm has a better removal of broken spot and noise. Comparing the result of interpretation of MODIS data and the result from the algorithm, we can conclude that the algorithm can better complete the MODIS image segmentation and on this basis we can have a better result of the sea ice edge.

Key words: agglomerative hierarchical clustering method, sea ice edge, MODIS, ice-water identification