Development of a mobile online detector for soil total nitrogen based on visible and short-wave near-infrared spectroscopy

Abstract

To develop a soil total nitrogen (STN) detection equipment with high accuracy, strong adaptability, and low cost, this study presents an innovative approach and successfully develops a mobile online STN detector based on visible and short-wave near-infrared (VIS-SWNIR) spectroscopy. The equipment mainly consists of optical system, mechanical system and control system. Different from most of the previous devices, it no longer relies on special optics. Instead, the equipment uses self-developed software to obtain the soil characteristic spectral data through real-time soil spectral extraction and invokes the STN model to obtain the detection value. A dataset comprising spectral reflectance and STN was constructed using 100 soil samples for model training. The Multiplicative Scatter Correction, First Derivative, and Competitive Adaptive Reweighted Sampling algorithms are selected to construct the PLSR model, based on the comparison of the modeling results from 10 preprocessing forms and 3 characteristic wavelength selection algorithms. The coefficient of determination for calibration and validation are 0.8767 and 0.8366, respectively. The Root Mean Square Error of calibration and validation are 0.0229 and 0.0298, respectively. To ensure the accuracy and stability of the equipment, appropriate tests are conducted on it in the laboratory. According to the results, during continuous operation for 2 h, the collection error of the detector’s characteristic wavelength reflectance remains within 1.80%. The self-developed software demonstrates sensitive responsiveness and strong stability. Moreover, the equipment exhibited strong performance in laboratory sample testing and field simulation tests, achieving correlation coefficients of 0.8862 and 0.8780, respectively, between measured and true values.