AdaBoost-SVR Model-Based Transmitted Microwave Sensing in Wheat Moisture Prediction
Dong Dai, Xu Mao, Yehong Liu, Du Chen, Shujin Guo, Shumao Wang, Bin Zhang
Abstract
The moisture content (MC) is a critical factor influencing the quality of wheat. Presently, the existing MC sensing methods do not provide a swift and precise determination of the wheat’s MC. This study develops a noncontact microwave sensing approach for the accurate prediction of wheat’s MC. To be more specific, the system employs microwave signals for transmitting wheat samples, enabling the measurement and calculation of the dielectric constant <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\varepsilon ^{\prime }$ </tex-math></inline-formula> and dielectric loss <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\varepsilon ^{\prime \prime }$ </tex-math></inline-formula> . Additionally, it takes into account the natural packing density of the sample, thereby enhancing detection accuracy. Meanwhile, the system receives the signal attenuation difference <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\Delta A$ </tex-math></inline-formula> and phase shift value <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\Delta \varphi $ </tex-math></inline-formula> corresponding to each frequency component, MC levels, and bulk density. Experimental results demonstrate that the wheat’s dielectric characteristic parameters monotonically increase with the MC level in the selected frequency range. According to the data feature, this study proposes an adaptive boosting for support vector regression (AdaBoost-SVR) moisture prediction model to consider dielectric characteristic parameters and bulk density comprehensively. These results are also compared with SVR regression and the linear regression predictions using only one dielectric parameter. The comparison denotes that the best prediction performance is obtained using the three feature inputs of dielectric characteristic indicators and the bulk density ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\varepsilon ^{\prime }$ </tex-math></inline-formula> & <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\varepsilon ^{\prime \prime },\rho $ </tex-math></inline-formula> ) achieving a root mean square error (RMSE) of 0.2201 and a coefficient of determination ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R}^{{{2}}}$ </tex-math></inline-formula> ) of 0.991. This study strongly suggests that a combination of multiple indicators is inevitable when predicting MC in stockpiled wheat or future silos.