Spectroscopic measurement of near-infrared soil pH parameters based on GhostNet-CBAM

Soil pH is an important parameter that affects plant nutrient uptake and biological activity and has received extensive attention and research. In this paper, we propose a neural network algorithm using Ghostnet combined with Convolutional Block Attention Module (CABM) to realize the near-infrared (NIR) PH spectral measurement of soil. The method firstly utilizes Monte Carlo Cross Validation (MCCV) method to reject the anomalous samples in the data, and then uses GhostNet combined with CBAM algorithm to train and predict the PH values of the four Lucas soil spectral data measured by the two different methods, and compares the prediction results with those of PLSR and VGGNet-16 methods. The results showed that the [Formula: see text] of the GhostNet-CBAM method could reach up to 0.9447, and the RMSE reached as low as 0.3267, and the scatter density plots of the predicted and true values further confirmed that the method could quickly and accurately obtain the soil pH parameters.