Detection Of Strong Convective Weather Based On Manifold Learning

Strong convective weather is one of the main meteorological disasters, so it is akey object that the meteorological department has been monitoring. To improve thedetection and forecast rate of strong convective weather effectively and quickly canreduce the loss of people’s lives and property and the production of industry andagriculture.As a new generation of weather radar, Doppler weather radar can provide us withrich image products, which can effectively improve the detection rate of strongconvective weather, but it brings a lot of high-dimensional data, which is difficult tobe used in weather forecasting. Therefore, data dimension reduction becomesimportant. As a nonlinear dimensionality reduction manifold learning has become ahot research field.The main work of this paper includes:(1) The24-dimensional database of the radar image feature was constructed.First, the base data of the Doppler radar was transformed into the bitmap. Second,through the image processing algorithm,24features of a monomer were extracted andthen written into the database, which can provide the base of data for the later work.(2) The dimensionality of the24-dimensional database of the radar image featurewas reduced by using manifold learning algorithm. In the low-dimensional database,the data was preprocessed, which was prepared for the later work of data mining.(3) In the low-dimensional database, the data was discretized in the way of equalfrequency, and then through data mining based on rough set theory, the knowledgebase for the detection of rainstorm and hailstone was formed. In the last, theeffectiveness of the knowledge base was assessed. The experimental results showedthat, through the manifold learning algorithm and data mining based on rough settheory, the system can effectively and quickly detect the area of rainstorm andhailstone in the radar image.To sum up, this paper opened up a new way to effectively and quickly detect thestrong convective weather, and to some extent, the results meet the need of practicalapplication. This paper provided a reference for the detection of the strong convectiveweather based on Doppler radar system.