Machine Learning for Better Food Series Article I: High Grade Olive Oil
Posted by Scintillon Institute on April 02, 2018
In the current age of information, where digitization is becoming indispensable (turning signals into digital format; i.e. “0s” and “1s”), photographs can be used as a great source of data. They can be transformed into mathematical databases, where each color of a single pixel can be seen as a set of three numbers representing the intensity of red, green, and blue channels. Therefore, the higher the number of pixels in a photo, which is directly related to its resolution, the greater the amount of information that can be used for different applications.
In this specific scenario, several photographs of olives, of different quality grades, were gathered, and their pixel maps were extracted through image processing. Afterwards, this information was used to train intelligent mathematical models to distinguish the olives in terms of quality grade. This intelligent modeling is also known as machine learning (computational artificial intelligence), which is becoming more and more popular within the scientific community. In many cases, its use is turning out to be a necessity, as it is the only way to process the immense databases that arise from fields such as food technology, biochemistry, or biomedicine.
Regarding the presented application to classify olives, the mathematical model employed is a type of artificial neural network (ANN). ANNs form a group of algorithms that acquired this name due to the fact that their design was initially inspired by the biological neuron and the mechanism it uses to learn and to transmit signals. To train the ANN, data from different olive images was inputted with the goal set to distinguish them according to their quality. Reaching accuracies of around 90% for the hardest cases, this mathematical tool proved to be useful to solve this problem (Pariente et al. 2018). Therefore, ANNs and machine learning have the potential of aiding olive oil producers to only use the raw materials (olives) that meet their quality standards, which will become the first step of the production of a high-quality olive oil.
This work was performed at the Complutense University of Madrid. This article is written by Dr. John Cancilla, a co-author of the original publication and currently a researcher at the Scintillon Institute in San Diego, California. The Scintillon Institute develops biological and computational technologies to find treatments for diseases and solutions to the environment.
Pariente ES, Cancilla JC, Wierzchos K, and Torrecilla JS. On-site images taken and processed to classify olives according to quality – The foundation of a high-grade olive oil. Postharvest Biology and Technology, 2018; 140; 60-6.