Enhancing greenhouse agriculture with IM-AlexNet: A novel method for accurate pest and disease identification

Abstract:China is the world's largest producer of greenhouse vegetables. Although greenhouse cultivation can extend the vegetable supply cycle and improve yield and quality, the closed environment can easily lead to a high incidence of pests and diseases, posing huge challenges to prevention and control work. To address this problem, this study proposes a greenhouse vegetable pest and disease identification method based on the improved AlexNet (IM-AlexNet) model. By introducing the ReLU6 activation function, batch normalization algorithm and GoogleNet Inception-v3 module, the model significantly improves the recognition performance and effectively solves problems such as noise interference, poor model convergence and target positioning overfitting. Experimental results show that the IM-AlexNet model is better than the traditional model in indicators such as Precision, Recall, F1 and MAP. Specifically, its MAP value is 88.91%, which is 10.77%, 8.6% and 5.14% higher than the AlexNet, CNN and YOLO-v7 models respectively. In addition, its ability to identify pest and disease targets in complex backgrounds is significantly enhanced, effectively reducing the missed detection rate, and showing stronger generalization capabilities under small sample conditions. This study shows that the IM-AlexNet model can help greenhouse vegetable growers quickly and accurately identify pests and diseases, thereby reducing the use of broad-spectrum pesticides, saving resources, and protecting the environment and consumer health. The results of this research provide an effective tool for intelligent monitoring of pests and diseases in greenhouse cultivation, and also provide an important reference for further research in related fields.