A Data-Driven Monitoring And Early Warning System Of Cucumber Downy Mildew In Greenhouse

Cucumber downy mildew is one of the main diseases that threat cucumbers cultivated in various methods all over the word.The disease spreads fast with strong epidemic and heavy damage.It is easy to cause losses to cucumber production.The relatively closed structural features of the solar greenhouse can easily meet the environmental requirements of cucumber downy mildew for medium temperature and high humidity.Therefore,timely monitoring and early warning of the occurrence of cucumber downy mildew can provide decision support for disease prevention management,and be of great significance for reducing losses,stabilizing production and ensuring quality.With the advancement of technology,the application of information technology has provided more and more effective methods for predicting plant diseases.Among them,machine learning,which extracts the inherent laws or patterns in the data to predict unknown data,does not need to consider the complicated mechanism process between various factors and the disease.Such data-driven models are more accurate than traditional empirical models.Compared with mathematical models,it is easier to establish the relationship between various pathogenic factors and disease occurrence,which provides a convenient and effective way for monitoring and early warning of cucumber downy mildew in greenhouse.It is convenient for cucumber growers to take timely decisions about the prevention and control of downy mildew.The main content of this article is following:(1)Determined the model input factor.A large number of relevant literatures at home and abroad were read and analyzed to understand the pathogenesis and main influencing factors of cucumber downy mildew.According to the research results of a large number of experts and scholars and the actual operational conditions,the input factors of the greenhouse cucumber downy mildew prediction model were determined as indoor nighttime average temperature,indoor nighttime average relative humidity,indoor daytime average temperature,indoor daytime average relative humidity,indoor cumulative hours of relative humidity?80%,outdoor daily maximum temperature,outdoor daily minimum temperature,outdoor daily average relative humidity and outdoor day average wind speed.(2)Data acquisition and model construction.The author collected environmental data and incidence of cucumber downy mildew in the greenhouse and environmental data outside the greenhouse of the experimental base.All data was collated and standardized combing with historical experimental data.The support vector machine and decision tree algorithm were used to construct the greenhouse cucumber downy mildew prediction model.The support vector machine algorithm used radial basis kernel function,linear kernel function and polynomial kernel function to construct the model.Used the above models to predict whether cucumber downy mildew would occur in the next three days.The prediction results of these four models were evaluated.It was found that prediction model of support vector machine with radial basis kernel function had the highest value in each evaluation standard,indicating that it worked best.(3)System design and realization.Based on the constructed predictive model,combined with MVC framework and Web Service technology,a data-driven web-based monitoring and early warning system of cucumber downy mildew in greenhouse was designed and developed.In order to facilitate the collection of greenhouse downy mildew disease information,an APP for pest information collection in mobile terminal was developed.The monitoring and early warning system includes monitoring and early warning of the onset of cucumber downy mildew in the greenhouse and management of various information in the cucumber planting base.The system could predict the onset time of downy mildew through the greenhouse cucumber downy mildew prediction model.Thereby the data-driven prediction was realized.Then used the real disease information obtained by APP to verify the prediction results,and stored the collected data for the construction of the downy mildew prediction model in the next cucumber planting.The model can be updated according to the data to form a circular data stream,which provides better decision support for the users.