Hyperspectral imaging techniques for rapid detection of nutrient content of hydroponically grown lettuce cultivars

Abstract

The analyses of online quality measurements of four lettuce cultivars (Rex, Tacitus, Black Seeded Simpson, Flandria) using hyperspectral image processing techniques have been studied. Seedlings were planted in Rockwool cubes and fed for 3 weeks using hydroponic nutrient solution containing 0, 50, 100, 150, 200, 250, and 300 ppm nitrogen. The hyperspectral images of the freshly cut lettuce leaves were captured using hyperspectral camera in the wavelength range of 400–1,000 nm. The algorithm to measure nutrient levels of leaf tissues including nitrate (NO3–), calcium (Ca2+), potassium (K+), soluble solid content (SSC), pH, and total chlorophyll concentration (SPAD reading) for lettuce was developed. To locate the optimum wave band combination values, simple linear regression models with two spectral data values, namely reflectance and its first derivative, and two spectral indices, namely ratio spectral index (RSI) and normalized difference spectral index (NDSI), were found. The optimal wave bands were within 506–601 nm and 634–701 nm. The nutrient levels of NO3–, Ca2+, K+, SSC, pH, and SPAD in the leaf samples were estimated using partial least squares regression (PLSR) and principal component analysis (PCA) techniques of the four optimal wave bands. Both PLSR and PCA models produced well correlated outcome in comparison with the laboratory measured freshly cut lettuce leaves in predicting nutrient contents with R2 = 0.784–0.987.