Improving the monitoring and early warning capability of catastrophic convective weather is the primary target of short-term weather forecasting and nowcasting. Convective precipitation/wind is highly uneven in time and space, and realtime monitoring is difficult. As for the traditional identification methods using infrared and water vapor brightness temperature for satellite remote sensing monitoring of precipitation, the number of alarm clouds is large, the false alarm rate is high, and the indicative significance is poor. It is necessary to find ways to extract more hidden information in satellite radiation observation combined with background factors and establish connections between the cloud top brightness temperature and the location of such disaster weather. This study attempts to use data of FY-2 infrared cloud imagery and intensified ground observation of hourly precipitation to classify and extract parameters by tracking cloud movement, and then use fuzzy support vector machine (FSVM) method to establish mathematical relations for machine learning between ground observation of rainfall intensity and dynamic evolution of cloud cluster. Based on the mathematical relationship, the cloud cluster and the heavy precipitation center can be identified with the significance of monitoring and early warning. This work also illustrates the possibility of developing satellite products for heavy precipitation cloud detection suitable for disaster monitoring and early warning in specific seasons.