An early warning model for vegetable pests based on multidimensional data

Abstract

Based on the research of sensor networks, pest monitoring equipment and systematic research, we mainly studied the major pests of south China vegetables, such as Bemisia Tabaci, Beetles, Plutellaxy Lostella and Thrips Tabaci. Using the multi-sensor network system, we collected the multi-dimensional information of the number of pests, soil, environment, ecological climate and meteorological factors in real time, on which a multi-dimensional big data based vegetable pest early warning model was constructed. According to the number of pests, the model used k-means algorithm to classify the damage into four levels: Mild, Moderate, Moderately Severe, and Severe. The correlation coefficient and gray relational degree were used to find out the key factors between the number of pests and the multi-dimensional information such as the vegetable soil, environment, ecology climate and meteorology, and based on BP neural network model, the key impact factors were trained, we find out five key impact factors: Rainfall Volume, Carbon Dioxide Concentration, Soil Temperature, Air Temperature and Foliar Humidity from June 2016 to February 2017. Finally, the selected characteristic data were normalized and then learned by BP Neural Network. The results showed that the recognition rate of the pests in southern vegetables was 96.7%. The algorithm is proved to be of high availability, meets the needs of early warning of pests and has a broad application prospect.