The classification of wheat species based on deep convolutional neural networks using scanning electron microscope (SEM) imaging.

With the increase in world population, decreased farmland, and global climate changes, the search for intelligent systems has become important to maintain product quality and productivity. For the continuous, quality, and clean production of wheat, which is a basic macronutrient for human health, it is necessary to use high-yield, high-quality, and unmixed clean seeds. The target of this review is to provide computer-aided identification of wheat species in the food industry and to offer taxonomists an opportunity to overcome classification difficulties. This research also highlights a systematic and functional wheat identification approach using scanning electron microscope (SEM) imaging techniques. The review found that not only accurately classifies the SEM images of wheat species of deep learning methods but also yields them being distinguished under different environmental conditions (irrigated and non-irrigated). Recently, the EfficientNet models are noteworthy as providing higher training speed and better parameter efficiency compared to the previous models. Therefore, the EfficientNet-B4 and EfficientNetV2-M were utilized on the top of an effective pre-processing task. Findings confirm that SEM imaging is outstanding when it comes to diagnosis and classification in the agricultural industry. The experimental results show that the proposed technique provides significantly better quantitative results and higher accuracy rates than the state-of-the-art Convolutional Neural Networks (CNN) algorithms. [ABSTRACT FROM AUTHOR]