A Miniaturized Spectral Module for the In-situ Detection of Layered Soil Moisture

The detection of soil moisture content along the depth of soil plays a crucial role in comprehensively obtaining the water deficiency conditions of crops. Existing methods for measuring soil moisture content cannot provide real-time data at centimeters-cale resolutions, which fails to meet the high-precision requirements of smart agriculture. This paper developed a miniaturized spectral module based on the AS7341 chip and established a soil moisture prediction model using the partial least squares regression (PLSR) approach. This system enables in-situ and online detection of layered soil moisture content within a depth of $0-15 \mathrm{~cm}$ and at a resolution of 1 cm. The results show that the Savitzky-Golay (SG) smoothing algorithm produces superior prediction performance compared to multiple scatter correction and first derivative preprocessing algorithms. The developed PLSR model achieved a coefficient of determination ($\mathrm{R}^{\mathbf{2}}$) of $\mathbf{0. 8 6 4}$ and a root mean square error of prediction (RMSEP) of 0.042.