Label-free non-invasive classification of rice seeds using optical coherence tomography assisted with deep neural network

Identification of the seed varieties is essential in quality control and high yield crop growth. The existing methods varietal identification rely primarily on visual examination DNA fingerprinting. Although pattern fingerprinting allows precise classification but fraught with challenges such as low rate polymorphism amongst closely related species, destructive method analysis a huge cost involved robust markers simple sequence repeat (SSR) single nucleotide polymorphisms. Here, we propose fast, non-contact non-invasive technique, deep learning assisted optical coherence tomography (OCT) for subsurface imaging order to distinguish different varieties. volumetric dataset of, (a) four rice (PUSA Basmati 1, PUSA 1509, 44 IR 64) and, (b) seven morphologically similar seeds landrace Pokkali was acquired using OCT technique. A feedforward neural network implemented feature extraction classify images into their relevant classes. proposed provides accuracy 89.6% total 158,421 82.5% classifying 56,301 collected from seeds. current technique can accurately irrespective morphological similarities be adopted removal duplication assessment purity