In this paper, the data of surface precipitation intensity of not less than 20 mm/h during eight regional heavy rains in Chongqing in 2021 are collected, and the multi-channel data of the Himawari-8 geostationary meteorological satellite in the corresponding time and space are matched to establish the one-to-one correspondence dataset between precipitation intensity and radiation features. On this basis, by using unit feature space classification analysis, multiple channels such as 0.64 m, 1.6 m, 3.9 m, 6.2 m, 7.3 m and 10.4 m are selected to form two-dimensional spectral space, and the distribution range as well as cluster characteristics of heavy precipitation pixels are retrieved. Based on the inversion results, the short-time heavy precipitation identification research in Chongqing is carried out. A threshold identification scheme of heavy precipitation areas based on satellite multi-channel data is also proposed in this paper. This scheme is used to test 16 cases of severe convective weather in Chongqing from 2021 to 2022. The results show that the scheme can realize the continuous dynamic monitoring of the heavy precipitation area 24 hours a day, 10 minutes by 10 minutes. The accuracy of the inversion of the precipitation area by the satellite data at the whole point, 10 minutes before the whole point and 30 minutes before the whole point can reach 70%－80%.