[Objective] To solve the low effici

[Objective] To solve the low efficiency and accuracy of maize leaf disease identification, a new maize disease identification method was proposed by combining the lightweight convolution neural network MobileNetV2 model with migration learning. [Method] Using the combination of traditional convolution neural network MobileNetV2 and migration learning, three training models were obtained by feature extraction, full migration and fine-tuning in migration learning, respectively, and compared with the newly trained MobileNetV2 model. [Result] It shows that fine-tuning model with less epoch experience can achieve higher accuracy and better recognition effect, and the model accuracy is 99.25%. The accuracy of the new training MobileNetV2 model was improved by 3.09%. Based on the test method, a maize disease identification system based on moving end was developed. The system test results show that the average recognition accuracy is 84%, and the time is only 1.16 seconds. [Conclusion] The system can identify maize leaf disease in the field environment.