基于VIS NIR的播种沟内土壤水分测量传感器研究  

Soil Moisture Measurement Sensor Research in Seeding Ditch Based on VIS NIR

作　　者：张东兴[1,2] 刘江 杨丽 崔涛[1,2] 和贤桃 张天亮[1,2] ZHANG Dongxing;LIU Jiang;YANG Li;CUI Tao;HE Xiantao;ZHANG Tianliang(College of Engineering,China Agricultural University,Beijing 100083,China;Key Laboratory of Soil Machine Plant System Technology,Ministry of Agriculture and Rural Affairs,Beijing 100083,China)

机构地区：[1]中国农业大学工学院,北京100083 [2]农业农村部土壤机器植物系统技术重点实验室,北京100083

出　　处：《农业机械学报》2021年第2期218-226,共9页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(32071915);国家玉米产业技术体系建设项目(CARS-02)。

摘　　要：基于土壤水分的播深调整技术,需要对播种沟土壤水分进行测量,以便根据落种点处的土壤水分信息进行播种调节,改变播种策略。本文设计了一种可见光近红外(Visible and near-infrared,VIS NIR)式土壤水分传感器。使用高分辨率光谱仪采集不同水分梯度的土壤光谱数据,采用偏最小二乘回归法(Partial least squares regression,PLSR)进行建模分析,并结合多种数据降维方法进行变量筛选,得出不同土壤含水率的敏感波段分别在410、540、780、970 nm附近;通过对这4种波长进行组合建模分析,选择得出预测最优的VIS和NIR波长组合为410 nm和970 nm。采用这两种波长设计传感器,并进行实验室试验,结果表明:当传感器与被测土壤表面距离d较近时(0~3 mm),测量精度和稳定性最好;当d为0~3 mm、土壤质量含水率处于0.69%~28.45%时,真实值与预测值之间决定系数R^(2)达到0.81,均方根误差(RMSE)为2.90%;当土壤质量含水率处于0.69%~22%时,真实值与预测值之间R 2提高至0.93,此时均方根误差降低为1.72%。通过析因试验得出,在显著性水平为0.05时,温度与光照强度对传感器正常工作没有明显影响。土槽试验表明,真实值与预测值之间R^(2)为0.82,RMSE为1.23%,满足玉米等作物播种环节土壤水分的测量要求。Soil moisture content(SMC)plays a vital role in seed germination and crop growth.It is of great significance for precision agriculture to acquire the SMC of seed-dropping point in planting for the sake of decision-making and depth-regulating of seeding.Thus,developing a proper SMC sensor will contribute a lot to precision agriculture.An SMC sensor was designed by using visible and near-infrared(VIS NIR)light source.The spectral data of soil samples was collected by a high-resolution spectrometer,then the partial least squares regression(PLSR)was used for determining the optimal pretreatment method,and various dimensionality reduction methods were employed to select the characteristic wavelengths of soil moisture.It was concluded that the sensitive reflectance bands of different SMC within 400~1000 nm were around 410 nm,540 nm,780 nm and 970 nm.Through the modeling analysis of combinations of two of these four wavelengths,the optimal wavelengths of VIS NIR light sources for prediction were selected as 410 nm and 970 nm,respectively.The results of experiments conducted in the laboratory showed that when the distance between the sensor and the measured soil surface was under 3 mm,within the range of 0.69%~28.45%SMC,the predicted and the measured values appeared a justified linear correlation for which the coefficient of determination(R^(2))was 0.81 while the root mean square error(RMSE)was 2.90%;within the range of 0.69%~22%SMC,the R^(2)of the linear model reached 0.93 and the RMSE was decreased to 1.72%.The factorial test indicated that temperature and light scarcely had influence on the SMC sensor at 0.05 level.The results of simulated field tests indicated that rocks and the process of acquire soil sampling may generate outliers.The R 2 of the linear correlation reached 0.82 and the RMSE was 1.23%after the outliers were excluded,which met the requirement of SMC detection in most conditions of precision agriculture such as maize planting.

关 键 词：可见光近红外 土壤含水率 传感器 精密播种 偏最小二乘回归 

分 类 号：S237[农业科学—农业机械化工程]