Search

The soil spectral characteristics are affected by soil particle size, water, and other factors, resulting in different spectral characteristics of soil in different regions due to different climate and geographical conditions. The soil composition prediction model established by using near-infrared spectroscopy technology often has its uniqueness. In this paper, in order to provide reference for the subsequent development of a soil nitrogen content rapid analyzer based on near-infrared spectroscopy technology, the soil nitrogen content was predicted by visible and near-infrared spectroscopy taking soil in Shanxi maize field as the research object. 120 soil samples were collected from the maize test field of Shanxi Agricultural University, and the total nitrogen content was determined and the visible and near-infrared spectroscopy was collected. The calibration set and prediction set were divided by the spectral physicochemical value co-occurrence distance(SPXY) algorithm at a ratio of 2:1. The effects of soil nitrogen content prediction models were analyzed, the models were established by combination with three modeling methods including partial least squares(PLS), support vector machine(SVM), and principal component regression(PCR), using six preprocessing methods including smoothing, standard state transformation, baseline correction, de trend, normalize, and multiple scattering correction(MSC), and the best prediction model was selected. The results showed that among the 21 prediction models established, the prediction model established by smoothing pre-processing method and PLS had the best effect, with the orientation coefficient(R2 ) of 0.907 and the prediction root mean square error(RMSEP) of 0.086. This prediction model could effectively predict the soil nitrogen content of maize fields in Shanxi province. [ABSTRACT FROM AUTHOR]