Research On Crop Disease Condition Monitoring And Early Warning Technology

In the process of crop growth,effectively detection of disease status and early warning are the premise to ensure the healthy growth of crops.At present,it is difficult to control the environmental factors of disease induction in the process of crop growth.At the same time,the disease early warning information lags behind,the early warning effect is poor,and the disease monitoring and identification algorithm is too complex to be realized on the resource limited embedded equipment.Therefore,Aiming at the disease prevention and control of greenhouse crops,The disease early warning scheme before the onset and the disease image recognition scheme after the onset are designed respectively,and the monitoring and early warning system is built to assist farmers in the comprehensive prevention and control of diseases and the daily management of greenhouse.The main research work of this paper includes:(1)Early warning scheme design of crop disease state.Firstly,the relationship between crop pathogenesis and environmental parameters is explored.Taking cucumber as an example,the infection conditions of the disease are clarified,and the changes of environmental information in the future are predicted by collecting the environmental information of greenhouse;Then,judge the disease trend of crops according to the environmental conditions of disease infection.Finally issue the disease classification early warning to assist farmers in greenhouse environmental regulation and control.In addition,the combined model method is adopted for the prediction of environmental parameters,and the moving average method(MA)is used to decompose the original environmental sequence into linear sequence and residual sequence,giving full play to the trend mining advantages of ARIMA in linear sequence and the nonlinear fitting ability of SVR for small amount of data respectively.By extracting the internal combined features of data,the accuracy and timeliness of prediction are improved.(2)Identification scheme design of crop disease state.In order to develop equipment to assist small farmers in disease image recognition.In this thesis,the residual structure and lightweight structure are introduced to design an efficient residual module and an identical residual module,and a lightweight convolutional neural network model is built based on the two modules.After testing,the lightweight convolutional neural network model not only ensures the accuracy of disease identification,but also reduces the requirements of the model for hardware computing resources.By analyzing the effect of disease identification under simple background and complex background,the applicability of the model in resource constrained equipment is verified.(3)Design and implementation of the system.the overall architecture is designed with the main controller and mobile phone as the core based on the completion of equipment selection and hardware connection of acquisition nodes,the client application is developed for environmental monitoring,disease early warning and disease image recognition.Combined with the server,the remote access function of the mobile terminal to the greenhouse is realized.(4)System testing and analysis.The monitoring and early warning system is built in the form of Wi Fi ad hoc network.With the main controller on site as the core,the communication performance of the whole system was tested after the communication protocol formulated.The average packet loss rate of the built LAN is less than 0.4%,and the mobile device can still receive the disease early warning information in the LAN in case of failure of communication equipment such as router.At the same time,it can assist farmers in photographing and identifying disease images,which provides a theoretical basis and Countermeasures for the comprehensive prevention and control of crop diseases.