This project aims to develop a methodology based on artificial vision techniques and statistical learning for the prediction of the type of seeds (species, variety, viability, and vigor). Therefore, allowing the identification in real-time of the highest quality seeds, building relevant knowledge for the implementation of tasks of planting new crops. In this way, we propose the development of an automatic and non-destructive methodology, which detects the best seeds for the establishment of smart crops. As a result, we will design a computational method based on artificial vision techniques and statistical learning, that optimizes the tasks of planting and storing native seeds.