There are 1 000 frosted and 1 000 non-frosted images of tea gardens were selected as training samples from nearly 20 000 images of tea gardens, and the tea garden frost model was established by using the EasyDL image recognition function of Baidu AI open platform in order to intelligently recognize the tea garden frosting. The accuracy rate, F1-score, precision rate and recall rate of the tea garden frost model were 99.2%, 99.2%, 99.3% and 99.1%, respectively. The model test results showed that the average recognition confidence of 20 frosted images was 95.51%, and the average recognition confidence of 20 non-frosted images was 99.99%. The average recognition confidence of 80 frosted images in the non-sample area (tea area in Jiangnan) was 97.53%, and the average recognition confidence of 80 non-frosted images was 95.11%.