Machine Learning techniques for detection of contaminated seeds of wheat

Nowadays, we have a big amount of regulations regarding the limitations in the consumption of different substances that can harm our health, enabling us to have a healthier lifestyle. We are also living in an era of a big amount of technological advances, with the advent of the computer and many Data Science algorithms and applications. In many countries, the control of the said substances does not come only in the usage of illicit drugs, but also in toxic substances found in regular food that can harm the human and animal health. The technological advance has taken too many steps inside the production process, but in some aspects, the mankind still uses archaic methods to work. In this project, we aim to unite both the necessity of a regulation and the automation of a process in order to generate a raise in the quality of life to the maximum number of persons possible. Applying several Machine Learning techniques to data collected through spectral images of wheat grains, we aim to provide a better and faster selection of the seeds, improving the food quality, meeting the European Union’s regulation for the contamination levels of cereals by Deoxinivalenol. After the application of the techniques, the main focus was to compare its results taking into account the liability of the data, by its metrics, in order to provide a good option to escalate its usage in large production chains, producing healthier food in less time, with automated steps.