Abstract
Reliable measurement of soil moisture is one of the primary requirements for agriculture, irrigation management, and the study of soil–water dynamics. To improve the accuracy of a smart soil moisture sensor beyond the manufacturer's calibration, there is a need for soil-specific recalibration and improved sensor models. This paper discusses the development of a semi-automatic virtual instrumentation system integrated with a standard gravimetric method for soil-specific recalibration of a relatively new, low-cost capacitive soil moisture sensor, SoilWatch 10, in loamy–sandy soil. Using dual sensor with confidence-weighted averaging sensor fusion method, variance in the raw sensor readings is minimized. Further, the pre-processed recalibration data is used to develop linear, polynomial, and optimized 3-layer FFBP-NN (feedforward backpropagation neural network) direct and inverse models of the sensor. Performance evaluation of the models indicate that goodness of fit and accuracy of the FBNN models is the highest, followed by the polynomial and linear models.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call