Research On Identification Method Of Clivia Scale Pests Based On Deep Learnin

Clivia is one of the important economic crops in China.It has high appreciation value because of its spreading leaves,evergreen green leaves and beautiful flowers and is often used to decorate courtyards.However,in the process of growth,it is easy to be infected by climate,geographical environment and pathogenic bacteria and germinate pests,and the main disease occurs in the leaves.Therefore,identification of pests and diseases by leaf changes is one of the main identification methods.In this paper,convolutional neural network is introduced to identify diseases and pests of Clivia,providing reference for the prevention and control of diseases and pests.The main research contents are as follows:(1)As there are few studies on the recognition of diseases and insect pests of Clivia and no public dataset,this paper constructs the disease and insect dataset through field shooting.Meanwhile,in order to reduce the risk of model overfitting caused by few samples,image enhancement algorithm is used to expand the dataset.Finally,the experiment shows that the m AP on the validation set reaches 91.73%,which is 17.62%higher than that before data enhancement,indicating the effectiveness of the image enhancement algorithm.(2)Aiming at the problem of poor recognition performance of current mainstream models on the dataset constructed in this paper,an improved YOLOX model for identifying diseases and pests of Clivia is proposed.Rep VGG network is used to replace the backbone network Darkenet-53 in YOLOX to improve the parallelism of the backbone network and reduce the latency.At the same time,the more efficient Vo VGSCSP module and the reparameterized Rep Conv convolution block are used to replace the CSPLayer module and the plain convolution module in the benchmark network PAFPN,so as to improve the network representation ability.Finally,compared with 10 mainstream target detection models,the m AP performance of the improved model is improved by 2.98%,and is significantly better than the existing mainstream models in other evaluation indicators.(3)According to the needs of flower farmers or greenhouse planting managers for disease and pest identification,a C# based disease and pest identification system for Clivia was designed and developed.The trained models are deployed into desktop programs to realize the application of algorithms in real production activities,helping users to accurately identify disease categories and disease severity,and providing technical support for accurate drug applications.