Identification of Nutrient Content of Psidium Guajava and Syzygium Cumini Leaves Using Hyperspectral Imaging

The Nutrient contents of the leaf play a vital role in the plant growth, metabolism, and nutritional values of the fruits, etc. However, the conventional method for identifying the leaf nutrition content is chemical analysis, which is timeconsuming, destructive, and labor-intensive. Therefore, a nondestructive approach technique, namely hyperspectral imaging, has been increasingly used to determine the spectral characteristics of plants. This study investigated hyperspectral imaging to identify the leaf nutrition content of two leaves (Psidium Guajava and Syzygium Cumini). To identify the nutrient content of the leaf, 30 samples were taken from the two leaves with four different conditions (standard leaf, diseased leaf, dried leaf, and pigmented leaf). Further, the spectral signature of the leaves was extracted and fed into the developed regression techniques. The three different types of developed regression algorithms such as Savitzky-Golay Partial Least Squares Regression (SG-PLS), Support Vector Machine Regression (SVMR), and Partial Least Squares Regression (PLS), were performed for the identification of nutrition content in the leaves. The results demonstrated that the SG-PLS could accurately predict leaf nutrition content using hyperspectral imaging data, with root mean squared errors RMSE =0.536 and $\mathrm{R}^{2}$=0.992 for Psidium Guajava and RMSE =1.54 and $\mathrm{R}^{2}$=0.936 for Syzygium cumini leaves in normal leaf condition. Further, the results show that hyperspectral imaging can be a powerful tool for nondestructive, rapid, and accurate measurement of plant nutrient status. Also, the proposed model will identify the difference between the normal, dried, pigmented, and diseased leaves using the RMSE values.