In order to carry out the research on cloud recognition and increase the objectivity of cloud observation, a cloud recognition method based on millimeter-wave cloud radar using neural network is presented. Through BP neural network algorithm ,12 cloud features of millimeter-wave cloud radar in Luoyuan, Pinghe and Jian’ou, Fujian Province, from August 2018 to June 2020 are used to recognize six types of cloud, including stratus, stratocumulus, altostratus, cumulus, altocumulus, cirrus. The recognition accuracy is about 78.13%, and the comprehensive evaluation index is about 79.84%. The research shows that this method can reduce the uncertainty of cloud observation and improve the accuracy of cloud automatic observation products under different weather backgrounds.