Research and predictive analysis of farmland leveling measurement devices

In order to accurately measure the farmland levelness, the measuring device used in this paper combines the indirect measurements by the ultrasonic sensor and direct measurements by magnetostrictive displacement sensor, obtains the measuring device posture-assisted correction measurement value by attitude sensor and gyroscope, and predicts the trend change of its measurement value by the different number of training sets of LSTM neural network. The test results show that the stability of the magnetostrictive displacement sensor is better than that of the ultrasonic sensor, and the noise can be effectively filtered by kalman distributed fusion data, and then through the first 10 s, the first 20 s, and the first 30 s data as training sets, respectively, to predict the analysis. The average root mean square error is 2.42 and the average absolute error is 2.67. The test results show that the root mean square error and the mean absolute error of the Kalman filter fusion data and the prediction data are small, which indicates that the measuring ddevice can accurately measure the farmland flatness and predict the changing trend.