关键词: 花生, 叶部病害识别, 注意力机制, 迁移学习, MobileNetV2

Abstract: Aiming at the problems of low identification efficiency of peanut leaf diseases and difficulty in one-site identification, a peanut leaf disease recognition algorithm based on the Convolution Block Attention Module (CBAM) and MobileNetV2 is proposed through transfer learning and by integrating the attention mechanism of CBAM. Firstly, a dataset of five kinds of peanut leaf disease images, including healthy leaves, black spot disease leaves, brown spot disease leaves, net spot disease leaves, and mosaic disease leaves, is established. Secondly, a peanut leaf disease recognition model is built by integrating channel attention mechanism and spatial attention mechanism. Finally, the recognition accuracy before and after model improvement is analyzed, compared with VGG16, InceptionV3 and ResNet50, and the detection time of a single image is predicted. The experimental results show that the accuracy rates of MobileNetV2, VGG16, InceptionV3 and ResNet50 are 97.54%, 97.34%, 96.06% and 74.88%, respectively, which are all lower than the accuracy rate of the improved model at 99.41%. The detection time for a single image is 0.061 s. The peanut leaf disease recognition algorithm based on transfer learning and attention mechanism is a lightweight neural network with high accuracy and few model parameters. It can be applied in peanut fields and used for on-site detection of peanut diseases by using mobile devices, enabling farmers to understand the growth status of peanuts in a timely manner.

Key words: peanut, leaf disease identification, attention mechanism, transfer learning, MobileNetV2