Research on Recognition Method of Zanthoxylum Armatum Rust Based on Deep Learning

This paper takes zanthoxylum armatum as the research object and studied a recognition method for recognition of the zanthoxylum armatum rust based on computer vision. First, we establish a leaf dataset with 22937 images, consisting of 19 kinds of leaf disease. Then we used deep learning method to analyze the disease of the crop leaf, and conducted 5 sets of experiments with different train set and test set ratio. The experiment results show that as the proportion of train set increases, the recognition accuracy of the model shows an upward trend. When the train set and test set ratio is at 8:2, the recognition accuracy reaching the best and it is 91.0%, which shows that the recognition model has good comprehensive performance and high practicability. The method used in this paper can recognize the rust of zanthoxylum armatum with a good performance, which has a positive effect on guiding agricultural production such as crop protection.