Rainstorm recognition based on similarity retrieval of rough set theory

For the urgent need of storm forecasting and warning, we achieved the rainstorm case retrieve system for the first time. We extracted the rainstorm radar image´s features from historical data set by using digital image processing technology, reduced the unwanted attributes, mined the minimum decision rules according to rough set theory, formed rainstorm knowledge base and case base, and achieved the forecast and recognition of strong convective weather finally. In this paper, the prediction medium scale was between 2km and 20km, the forecast aging was between 0 and 3 hours, and the rainfall amount exceeded 20mm during 3 hours. Experimental tests show that the accuracy of the rainstorm forecasting recognition is 87%, false alarm rate is 13%, alarm failure rate is 0, which meet the need of practical application and help people to make more accurate forecast. By the rainstorm case retrieve system, we can determine the similarity between the target case and the historical case and improve the knowledge base and the system´s intelligence.