Prediction and monitoring model for farmland environmental system using soil sensor and neural network algorithm

Abstract In this study, data fusion algorithm is used to classify the soil species and calibrate humidity sensor, by using edge computing a wireless sensor network, farmland environment monitoring system with two-stage calibration function of frequency domain reflectometer (FDR) established. Edge in nodes, including saturation value calculated hardness, calculation process neural model classification. A bagged tree adopted avoid over-fitting reduce prediction variance decision tree. established on each training set, C4.5 construct After primary calibration, root mean squared error (RMSE) between measured standard values reduced less than 0.0849%. The (MSE) absolute (MAE) are 0.7208 0.6929%. backpropagation network train dynamic dataset. output trained closer actual that FDR sensor. MAE, MSE, RMSE decrease 1.37%, 3.79, 1.86%. With accurate measurements humidity, research shows an important guiding significance for improving utilization efficiency agricultural water, saving formulating crop irrigation process.