Research On Intelligent Diagnosis And Identification System For Soybean Pests And Diseases

As one of China’s important cash crops,the gradual increase in soybean production and the growing acreage in China over the last decade has led to a sharp increase in the incidence of pests and diseases,which has become a key factor affecting soybean production in China.At present,the identification and diagnosis of soybean pests and diseases in China still relies mainly on traditional manual detection,which consumes a lot of manpower and lacks real time,and due to the subjective nature of the judgement,deviations often occur and effective control measures cannot be taken in the shortest possible time,thus failing to meet the development needs of today’s modern agriculture.With the booming development of modern agriculture,traditional manual diagnosis is gradually being replaced by intelligent diagnosis.In order to further enhance the effectiveness of soybean pest control and promote the development of smart agriculture,this thesis conducts in-depth research on expert systems and deep learning technologies to build an intelligent diagnosis and identification system to provide assistance in the diagnosis and identification of soybean pests and diseases,as follows:First,intelligent diagnosis of soybean pests and diseases is achieved based on expert systems.Knowledge is acquired in three ways: manual acquisition method,semi-automatic acquisition method and automatic acquisition method,and the knowledge base is designed using generative rules.An uncertainty-based reasoning method is used to determine the damaged parts of soybean pests and diseases,such as roots,stems and leaves,based on their pathogenesis characteristics.Based on the corresponding representations of different parts combined with the experience of domain experts,an expert system reasoning machine was designed and implemented to achieve assisted diagnosis of soybean pests and diseases.Secondly,intelligent identification of soybean pests and diseases was achieved based on convolutional neural networks.Two ways were used to obtain the experimental data,firstly,the web public data set;secondly,through web crawlers.Five categories of bacterial leaf spot,bean rust,heart-eating insect,bean aspergillus and health were selected as soybean pests and diseases as research objects.The soybean pest and disease dataset was pre-processed by horizontal flip,rotation and brightness adjustment,and the image size was uniformly adjusted to 224*224.Based on the ResNet34 network model,the soybean pest and disease images were trained by migration learning through modifying the model framework and freezing part of the layer model.Through experimental comparison tests,the results show that the model can be effectively used for soybean pest and disease identification with an accuracy rate of 93.5%,which can achieve pest and disease identification and classification more efficiently.Finally,an intelligent diagnosis and identification system for soybean pests and diseases was implemented.The system is built on the basis of expert system assisted diagnosis and intelligent identification mechanism based on deep learning.The main functions of the system include soybean pest and disease enquiry,pest and disease intelligent diagnosis,pest and disease intelligent identification and expert consultation.Through preliminary application,the research results can provide effective assistance in the diagnosis and identification of pests and diseases in the soybean cultivation process,thus enhancing the intelligence of soybean cultivation.