| As a large population country,China attaches great importance to the development of agriculture.The development of organic agriculture is the key problem to promote the development of information agriculture.It conforms to the concept of scientific development concept and meets the needs of people for green food in the period of industry 4.0 and rural urbanization.Pest control and early warning is an important step to promote the development of organic agriculture.At present,there are two kinds of pest warning methods:one is to record the number of pests for months or years by recording and analyzing the trend of pest occurrence,and the other is to set up an algorithm model to predict the degree of pest occurrence.On the basis of manual statistics,the occurrence of insect pests is pre-predicted by experience.The method is not suitable for the nonlinear characteristics of pest occurrence system.Therefore,how to establish the algorithm model and forecast the pest is not only the work of the professional organization,but also a kind of national demand.In this paper,an intelligent pest control early warning system based on neural network is designed.The system collects pest and temperature and humidity data through the data acquisition end,sends the data to the data sink by wireless transmission,preprocesses the data and forecasts the algorithm,and realizes the prediction of the pest grade.The research and analysis show that the pest occurrence system has the characteristics of uncertainty,timeliness and variability,which shows obvious nonlinear characteristics,and neural network is the main method to solve the multi-factor and complex nonlinear system at present.Based on this,the principle of BP pest prediction algorithm is studied,and the prediction model of BP pest occurrence degree is established,and the prediction of insect disaster grade is realized.Because BP algorithm is easy to fall into local minimum,genetic algorithm has the advantage of seeking optimization.A GA-BP combinatorial optimization algorithm is proposed to optimize the initial position of weight and threshold of BP neural network by genetic algorithm.The number of training and the amount of calculation are reduced,and the local optimization is not caused.The simulation results show that compared with the basic BP algorithm,the GA-BP algorithm effectively improves the accuracy of pest occurrence prediction.Based on the algorithm,an embedded platform based on FreeRTOS real-time operating system is built.Based on the idea of energy saving and green,the system is based on ARM microprocessor as the main controller and solar power supply mode.The pest data acquisition module,pest data convergence module and solar power supply module are designed.Data acquisition module,through temperature and humidity acquisition circuit,insect control counting circuit to collect pest data,as algorithm data samples;The data aggregation module,as the algorithn processing platform of pest data,mainly completes the function of pest prediction algorithn,and realizes the effective early warning of insect pests.The experimental results show that the accuracy of combined GA-BP and BP algorithm for pest degree prediction is 83.3%and 75%,respectively,and the prediction accuracy of combined GA-BP algorithm is increased by 8.3%on average.The training time of combined GA-BP and BP algorithn for pest degree prediction was 10.3s and 18.1s,respectively,and the prediction training time of the algorithm was shortened by 7.8 s.The acceptability of combined GA-BP and BP algorithl for pest prediction is 83.3%and 75%,respectively,and the acceptability of the algorithm is increased by 8.3%on average,which shows that the combined algorithl is effective for pest forecasting.Will combine GA-BP algorithm is applied to embedded platform,and the accuracy of pest prediction can reach 83.3%.It shows that the algorithm can be effectively applied to this platform,and realizes the early warning of pest occurrence and remote monitoring of pest information. |
PDF Full Text Request