Identification And Control Of Pests And Diseases In Green Ecological Peach Orchard Based On Deep Learning

Peach is a common fruit in people’s life,and it was also one of the important economic crops in my country.It will be affected by various pests and diseases in its production process.Therefore,it has important meaning for peach orchard production to be able to obtain,quickly identify and scientifically control peach orchard pests.With the development of artificial intelligence technology,image recognition and target detection have been gradually applied to many fields such as industry,military,commerce,medical treatment,etc.Using image processing technology to realize the diagnosis of agricultural diseases and insect pests has become a research hotspot of agricultural informatization.In particular,agricultural pests and diseases generally have the characteristics of small targets,but traditional target recognition algorithms often have problems such as the need to improve the recognition accuracy for the recognition of such small targets.The YOLO v5 s network model saves the image preprocessing work of manual feature selection and data mining,and also has an excellent recognition effect on the peach aphid whose body color is similar to that of peach leaves,which makes the YOLO v5 s network model useful for small targets such as agricultural diseases and insect pests.Recognition has great advantages.However,with the optimization of the YOLO network model,while its recognition accuracy has gradually improved,the parameters generated by training have become larger and larger,and such large-scale parameters have brought a lot of inconvenience to practical engineering applications.Therefore,there are many problems in simply using the YOLO v5 s model to detect agricultural pests and diseases.Based on the above analysis,the YOLO v5 s network is studied,combined with the contours,colors and other characteristics of pests and diseases in peach orchard,to design and study a more targeted and lightweight model to improve the application performance of the studied model in actual scenarios.The main work includes:（1）A peach aphid dataset was established,and images of peach aphid in the natural environment were collected through multi-angle and multi-distance shooting strategies.After filtering out invalid data,data enhancement and other methods are used to expand the data,complete the image labeling work,and provide data support for image recognition research.（2）The Ghost module and the Deep convolutional layer are introduced into the YOLO v5 s network structure,and the image recognition model of peach aphid is constructed combined with the attention mechanism.Linear operation is performed on the acquired intrinsic feature map of the peach aphid image,and then the output results are spliced to obtain a lightweight YOLO v5 s network model.After training,the peach aphid can be accurately identified,which provides a basis for precise control.（3）The green control system of peach aphid was introduced and verified by experiments.Relying on the improved YOLO v5s-based green peach aphid identification method,the Pytorch model trained in Python is called for peach aphid identification,and the peach aphid is controlled by ozone mist,which reduces the amount of chemical pesticides and reduces environmental pollution.Improve food safety.