Employing Nondestructive Approach of Spectral Imaging to Detect Artificially Degreened Lemon

The demand for reliable methods to detect artificially degreened citrus fruits is growing in the agricultural sector. In this letter, we propose a spectral imaging-based approach to differentiate natural and artificially degreened lemons using eight narrow spectral bands within the visible and near-infrared range. To support this research, we introduce the Spectral Imaging Lemon database, consisting of 7168 images of natural and degreened lemons. Experiments were conducted across the wavelengths from 530 to 1000 nm, leveraging six feature descriptors and a support vector machine (SVM) classifier. The proposed method achieved an impressive 93.5% average classification accuracy, showcasing its effectiveness.