Potential use of Hyperspectral Imaging for Authentication of Rocket Leaves according to Agricultural practices

In this study we assessed the potentiality of hyperspectral imaging for the discrimination of different management systems for water and fertilizer use on rocket leaves. Soilless cultivation of rocket leaves was conducted in unheated greenhouse testing a factorial combination of two fertilization levels (defined as ‘high’ or ‘low’, with reference to the standard dosage range recommended for the specific fertilizers used in the study) with two irrigation scheduling approaches (sensor-based, as a promising technique for optimal water supply based on real plant needs, and empirically timer-based, as an approach potentially resulting in un-efficient watering). Reflectance spectra were acquired using Vis-NIR ranges between400-1000nm and NIR ranges between900-1700nm. After pretreatment spectra were used for discriminant models of the four treatments and for ANOVA-simultaneous component analysis (ASCA) in order to understand the effect of each factor on spectral response. Comparing different treatments, PLS-DA model yielded the accuracy of 98.19%, 97.6%, and 97.2% for the cross validation, calibration and prediction system, respectively in Vis-NIR ranges while in NIR ranges the accuracy improved to 100%, 99.8%, and 99.5%. Moreover, applying ASCA significant wavelengths were selected. Results indicated promising potentiality of hyperspectral imaging for the authentication of products from low input managed agricultural systems.