Intelligent detection of potassium oxide in potassium sulfate fertilizer using hyperspectral imaging and artificial neural networks

Determining potassium oxide in chemical fertilizers is essential for maintaining soil health. Due to difficulty, time-consuming, high cost, and requiring chemical agents of laboratory methods, hyperspectral imaging was applied to predict potassium oxide in potassium sulfate based on its visible and near-infrared reflectance. After acquiring the hyperspectral images, 453.32, 669.95, 778.19, 891.55, 663.34, 847.72, and 869.32 nm were identified as the effective wavelengths. Subsequently, mean, minimum, maximum, variance, median, and standard deviation features were extracted and subjected to classification. Finally, the correlation coefficient of the prediction model based on the artificial neural network method was obtained as 0.92. The results suggest applying hyperspectral imaging to assess the quality of potassium fertilizer.