| At present,the identification of agricultural diseases mostly depends on the observation of human eyes by professionals and the judgment of artificial experience,which leads to problems such as misdiagnosis and slow speed of treatment,which brings a lot of economic losses.Therefore,the most critical problem of disease management is how to quickly identify the disease types,and then make scientific and effective response measures.With the rapid development of computer vision technology,traditional machine learning and deep learning methods have been widely used in the classification and recognition of crop diseases.Traditional machine learning for different diseases requires different feature extraction by hand,need a lot of work,while the convolutional neural network without feature extraction by hand,the advantages will greatly reduce the workload,but the existing leaf disease recognition model based on deep learning there is still a problem of large quantities and low identification accuracy.To solve these problems,some improvements were made on the existing convolutional neural network model for the identification of leaf diseases.The model preprocesses the collected images to ensure the consistency of the data.The processed image is analyzed by the optimized convolutional neural network and simulated.The results show that the optimized model has a great improvement in the accuracy of blade disease identification.Finally,in order to make it more suitable for practical application in blade disease identification,a related human-computer interaction interface was designed based on Qt Designer software,so as to realize the visual operation of the system.The main experimental contents and research innovations include the following aspects:(1)In order to improve the effectiveness and accuracy of the neural network model,the image training set needs to be preprocessed,such as image flip,adding salt and pepper noise and Gaussian blur,and the resolution of the image is unified to 100×100.Design of neural network model structure: Based on the classical VGG-16 model,a convolutional neural network composed of five groups of convolution-pooling layers and three global mean pooling layers is designed,and features are fused with the high-dimensional features extracted from Inception V2 structure in Goog Le Net model.In addition,in order to further improve the accuracy of the model,a new activation function,ELU,and batch normalization layer are also introduced.After setting the relevant parameters of the model,the Paddle Paddle framework was used to conduct comparative experiments on several classical models and the model in this paper under the same conditions.The results show that the recognition accuracy of the optimized model is higher than that of other models,and the recognition accuracy reaches 99.66%.(2)UI interface design based on Py Qt5: In order to improve the usability and reliability of the system,this paper designs and develops the interactive interface based on the design software Py Qt5 and Qt Designer developed by Python.Through the image of disinfected leaves uploaded by users to detect and identify quickly,it provides the relevant agricultural personnel with effective information,so that users can make effective prevention and control strategies in the actual production.In this thesis,the convolutional neural network algorithm in deep learning technology was used to design a timely and effective identification system for crop leaf diseases.Combined with the visual operation interface,the model was effectively improved in practice,and the problems of disease prevention and control encountered by the vast number of farmers in production were better alleviated. |
