Abstract
In China, many greenhouse farmers still rely on manual identification of vegetable pests and diseases. This method relies on traditional experience and intuitive observation, lacks scientific and systematic methods, which is prone to overuse in subsequent use of chemical pesticides. To this end, this study proposes a method for identifying greenhouse vegetable pests and diseases based on the improved AlexNet model. It uses the AlexNet as an image recognition model for pests and diseases and uses the ReLU6 activation function to solve the problems of poor model convergence and overfitting. It also integrates the GoogleNet Inception-v3 module to improve recognition results, which solves some problems of the AlexNet model, such as noise, poor model convergence, and over-fitting in target positioning. After compare with AlexNet, CNN (Convolutional Neural Networks), and YOLO-V3 (You Only Look Once Version 3.0) model, the IM-AlexNet model is superior to the other three models in MAP value, recognition accuracy, and loss function. It shows that the monitoring network designed in this study can better identify vegetable pests and diseases efficiently. It can help vegetable greenhouse farmers accurately and quickly identify vegetable pests and diseases, reduce the use of broad-spectrum pesticides, and save time and resources, which is beneficial to the environment and consumer health.