天气图相似检索研究

Similarity Retrieval of Synoptic Charts

天气图中含有丰富的气象信息,这些信息的载体具有图形和数值两种不同的表现形式.针对图形类信息,构建一种特征向量,该向量突出图形特征,利于相似测度,适合于展开基于内容的检索.以此为基础,将天气图的相似检索分为基于各气象要素的检索和相似综合评判两个阶段,能够满足预报业务的不同需求.另外知识库的使用使检索系统具有较小的冗余和较强的柔性.

The synoptic charts embrace much meteorological information, which can be represented in two forms: numerical values and curves. For curves, a new way to construct feature vectors is put forward in this paper. These vectors give prominence to the features of curves and facilitate the similarity measurement. The way is suitable for content-based retrieval,too. In the paper the similarity retrieval of synoptic charts consists of two stages. The first stage is the retrieval based on meteorological elements in the synoptic chart and the other is the synthetic evaluation of similarity. The framework can satisfy various demands in forecasting. In addition, the use of knowledge base can reduce redundancy and enhance the flexibility of the system.