Research On The Feature Extraction And The Construction Of The Feature System Of Developing Typhoon

The work of this paper belongs to the open project, "the research on the recognition method of typhoon during its developing stage", which is supported by the National Laboratory of Pattern Recognition.Typhoon is one of the most dangerous disaster climates, which heavily affect routine work and daily life. A good many attentions is paid to the typhoon forecasting, especially to its early-stage forecasting. During its developing stage, typhoon hasn't its typical features, which makes it very hard to recognize the typhoon in such stage. Therefore, this project enjoys a highly research value.In this paper, the differences between developing typhoon and non-typhoon are highly regarded. In order to describe and quantify these differences, we extract features from the developing typhoon and then construct the feature system in terms of the time series satellite images.Several static features are extracted according to the characteristics of developing typhoon, such as area, average gray value, circle-like ratio, key skeleton slope, and so on. Also, the temporal variation feature of each feature is extracted.A dynamic rotation feature, which is based on the cloud motion wind, is put forward for the first time. According to the test results, its effect is highly satisfactory, which will make a significant contribution to the accurate recognition of the developing typhoon in the near future.Based on the extracted features, both static and dynamic, and their temporal variation features, a set of recognition rules and a feature system are constructed, which provides the foundation for the construction of the typhoon forecasting and recognition model and the satellite image database.