基于电容法和深度补偿的机载式玉米播种种沟土壤墒情在线检测系统设计与试验

Design and Experiment of Airborne On-line Soil Moisture Detection System Based on Capacitance Method and Depth Compensation

摘要:

针对现有玉米播种装备缺乏土壤墒情在线检测系统，且已有的土壤墒情在线检测系统检测精度不高、环境适应差的问题，本文提出了一种基于电容法和深度补偿的土壤墒情检测方法，开发了一套机载式玉米播种种沟墒情在线检测系统。开展了电容器结构优化参数仿真试验，确定了最佳电容极板参数为：极板间距75.8mm、极板厚度0.7mm、极板相对面积5073mm2，其长度为100mm，宽度为50.73mm；系统硬件部分主要包括FDC2214型电容传感器、F4046型压力传感器和STM32F103型单片机，电容传感器用于获取待测土壤电容，压力传感器用于获取待测土壤压力，间接反推待测区域土壤深度；系统软件则利用Matlab平台进行开发，用于对土壤电容信号和压力信号的实时采集、计算、显示与存储。基于该系统探究了土壤墒情检测模型影响因素，构建了基于BP神经网络的土壤墒情检测模型，建模试验结果表明，当土壤墒情为7.23%～21.14%时，模型预测性能指标R2、RMSE和RPD分别为0.927、0.008和3.70，预测效果较好。最终，将构建的模型集成到土壤墒情在线检测系统，开展了台架与田间验证试验。台架试验结果表明，土壤墒情实际值与检测值的拟合决定系数R2均为0.852～0.927，土壤墒情预测结果绝对误差为-2.89%～2.57%，绝对误差平均值为1.01%；田间试验结果表明，土壤墒情检测值与实际值拟合曲线决定系数R2为0.842，土壤墒情检测绝对误差为-0.96%～0.45%，平均绝对误差为0.39%。本研究所研制的检测系统性能满足玉米播种机田间作业时土壤墒情检测需求。

Abstract:

Aiming at the lack of online soil moisture detection system for existing corn sowing equipment and the problems of low detection accuracy and poor environmental adaptation of the existing online soil moisture detection system, a method of soil moisture measurement was presented based on capacitance method and depth compensation, and a set of airborne corn sowing soil moisture online detection system was developed. In terms of electrode plate optimization, the system conducted simulation experiments on capacitor structure optimization parameters with electrode plate spacing, electrode plate thickness, and relative area as experimental factors. The optimal capacitor electrode plate parameters were determined to be as follows: electrode plate spacing of 75.8mm, electrode plate thickness of 0.7mm, electrode plate relative area of 5.073mm2, length of 100mm, width of 50.73mm. The hardware part of the system mainly included FDC2214 capacitive sensor, F4046 pressure sensor, and STM32F103 microcontroller. The capacitive sensor was used to obtain the capacitance value of the soil to be tested, and the pressure sensor was used to obtain the pressure value of the soil to be tested, indirectly inferring the soil depth in the tested area. The system software was developed by using Matlab platform for real-time acquisition, calculation, display, and storage of soil capacitance signals and pressure signals. Based on this system, the influencing factors of soil moisture detection models were explored, and a soil moisture detection model based on BP neural network was constructed. The modeling experiment results showed that when the soil moisture was in the range of 7.23% to 21.14%, the model’s predictive performance indicators R2, RMSE, and RPD were 0.927, 0.008, and 3.70, respectively, with good predictive performance. Finally, the constructed model was integrated into the online soil moisture detection system and bench and field validation experiments were conducted. The results of bench test showed that the fitting coefficient R2 of soil moisture content was 0.852~0.927. The absolute error range of soil moisture prediction results was from -2.89% to 2.57%, and the average absolute error was 1.01%. The field test results showed that the coefficient of determination R2 of the fitting curve between the soil moisture monitoring value and the actual value was 0.842, and the absolute error range of soil moisture monitoring was from -0.96% to 0.45%, with an average absolute error of 0.39%. This indicated that the performance of the detection system developed met the needs of soil moisture monitoring during field operations of corn seeders.