Diseases and Pests Identification of Lycium Barbarum Using SE-MobileNet V2 Algorithm

The prevention and control of crop diseases and insect pests is a key link in agricultural production. Accurate image recognition of diseases and insect pests is significant in agriculture and it has high efficiency, low cost and convenient operation. The traditional recognition method is difficult to recognize the pests relying on manual recognition, and it is easy to misjudge the pests. MobileNet V2 is an important algorithm for image recognition, which has high recognition accuracy and much lower parameter quantity than other algorithms. [1] This paper proposes a new SE-MobileNet V2 target recognition algorithm which improves the MobileNet V2 algorithm. The proposed algorithm embeds SE module between the last pooling layer and convolution layer of MobileNet V2 network, enhancing the receptive field of the network. It combines the advantages of MobileNet V2 and SENet, is more accurate than MobileNet V2 and is more suitable to recognize small target. In this paper, experiments are conducted based on data sets containing 11 kinds of lycium barbarum diseases and pests. Through experimental verification, the proposed algorithm has the highest recognition accuracy and strongest robustness than other algorithm. SE-MobileNet V2 target recognition algorithm improves the accuracy and speed of image recognition, is meaningful for accurate control of lycium barbarum diseases and pests.