Farmland Pest Monitoring Platform Based On Raspberry Pi

China is a big agricultural country,and agriculture plays an important role in the Chinese economy.In agricultural production,pests and diseases are important production bottlenecks,which seriously affect crop production.Therefore,pest management is crucial to the sustainable development of Chinese agriculture.Through the monitoring of pests and diseases in farmland,we can keep abreast of the number,distribution,danger and scope of pests and diseases,so as to formulate corresponding control programs and take targeted control measures to improve the control effect and reduce economic losses.This can effectively improve the yield and quality of crops,ensure the production and life of farmers,reduce the use of pesticides,and achieve sustainable agricultural development.With the development of information technology,the technology of farmland pest monitoring has begun to take shape.In order to realize real-time monitoring and early warning of crop pests data,including image data.Object detection is an important image processing technique that can be used to identify and locate pest and disease targets in crop images.However,the traditional target detection is faced with high network model complexity and time-consuming in the monitoring of farmland diseases and insect pests,resulting in poor detection effect.With the development of deep learning theory and technology,especially the introduction of lightweight models,this paper proposes the use of lightweight optimization algorithms to monitor farmland pests.By reducing the complexity of the network model and improving the operating efficiency of the model,thereby improving the detection speed of the model enables better pest monitoring and identification in crop images.The specific work is as follows:This paper proposes a lightweight detection method Retina Animal for pest detection in farmland,using the attribute pyramid module to distinguish simple targets by extracting shallow features and complex targets by extracting deep features.Use top-level features for precise classification,and low-level high-resolution features for accurate positioning.And using a single-stage state head module,the final feature extraction structure is obtained by merging three channels.Since the single-stage state header module used in this paper is a lightweight model,it has the advantages of small memory space and easy transplantation on mobile devices or embedded devices.In this paper,an algorithm based on edge computing is applied to pest monitoring in farmland.This method can not only reduce the impact of network delay,but also reduce storage costs and data transmission.In order to realize real-time monitoring of farmland pests,we deploy a lightweight model on the edge device Raspberry Pi,perform local computing and data storage on the edge device,and use the edge device to collect and process data.Under the action of the edge module neural rod,model reasoning is accelerated and efficient neural network calculations are performed.From the analysis of the experimental results,the algorithm proposed in this paper can effectively improve the monitoring efficiency of farmland pests,and has positive significance for increasing the increase of farmland crop production.