Detection of nitrogen content in wheat leaves based on visible/near-infrared spectroscopy sensing

Nitrogen is an important element present in vital substances such as plant proteins and chlorophyll, the content of which directly reflects the nutrient status of crops, and provides a theoretical basis for crop nutrient diagnosis, growth monitoring and yield potential prediction. Taking the chip-level visible/near-infrared spectral sensor AS7263 as the data acquisition module and the Arduino Uno single-chip microcomputer development board as the control module, a portable crop leaf spectral sensing system was designed in this study. The spectral reflectance and nitrogen content of wheat leaves were obtained through field experiments. Principal Component Analysis (PCA) was used to eliminate abnormal spectral data. Combined with pretreatment algorithms including Multiplicative Scatter Correction (MSC) and Standard Normal Variate (SNV), the prediction models for wheat leaf nitrogen content were established based on Partial Least Squares (PLS), Support Vector Machine (SVR), Random Forest (RF) and a Back Propagation (BP) neural network. The results showed that compared with SNV, the model performance based on the spectral data after MSC pretreatment was better. The test set R² values of PLS, SVR, RF and BP models were 0.61, 0.75, 0.83, and 0.89, and the root mean square errors (RMSEs) were 4.62 mg g^-1, 4.38 mg g^-1, 3.39 mg g^-1 and 3.27 mg g^-1, respectively. The MSC-BP prediction performance was the best, and the non-destructive and accurate detection of nitrogen in wheat leaves was realized, which verified the feasibility of micro-spectral sensing technology in crop nutrition diagnosis.