Research on vehicle-mounted soil electrical conductivity and moisture content detection system based on current–voltage six-terminal method and spectroscopy

To provide real-time, online, and accurate measurements of soil electrical conductivity (EC) and moisture content (SM) in large-scale agricultural fields and guide field management, vehicle-mounted online EC and SM measurement system based on the fusion of current-voltage six-terminal method and near-infrared spectral in-formation was developed and the predictive accuracy of the system was tested in field and laboratory experi-ments. The traditional current-voltage six-terminal method is used to fuse the spectral information at the moisture-sensitive wavelength of 1450 nm to predict SM and then a correction factor Q was proposed to elim-inate the influence of SM on EC. In addition, the linear relationship between soil apparent electrical conductivity (or bulk soil electrical conductivity, ECa) and SM, the exponential relationship between SM and spectral reflectance (R) were used to indirectly obtain the exponential relationship between ECa and R. And then, the binary model of EC is established using the electrical conductivity values of the univariate model (ECeq) obtained from the current-voltage six-terminal method and the univariate model (EClq) obtained from the relationship between the conductivity of the ECa and the R index as an input variable. The experimental data of the univariate model EClq and ECeq and the binary model were analyzed. The field electrical conductivity detection accuracy R2 of the univariate model EClq and ECeq is 0.790 and 0.829 respectively, and the RMSE is 10.113 and 8.078 respectively. While the field electrical conductivity detection accuracy R2 and RMSE of the binary model are 0.899 and 6.535 respectively. The laboratory electrical conductivity detection accuracy of R2 for the univariate models EClq and ECeq were 0.874 and 0.893, and the RMSE were 7.129 and 6.663 respectively. The laboratory electrical conductivity detection accuracy of R2 and RMSE for the binary model were 0.942 and 5.346 respec-tively. From the above result, it can be seen that R2 for the binary model has improved and the RMSE of the model has decreased. Analysis result of the SM prediction model shows that the accuracy of the SM prediction model R2 in the field and laboratory is 0.888 and 0.938 respectively, and the RMSE is 0.339 and 0.239 respectively, which indicates that the accuracy of the SM prediction model can meet the system requirements. The overall experimental results show that the electrical conductivity detection accuracy of the binary model of the detection system fused with spectral data is significantly improved, and the SM prediction accuracy can meet the needs of farmland detection, which can guide scientific and precise management of farmland.