• Artificial Intelligence ·
Return

Success case

Detection and

classification of olives

in the

oil production chain

The challenge

IRTA is a research institute of the Government of Catalonia, which seeks to promote research and technological development in the agrifood sector, as well as to contribute to the modernization, improvement and promotion of competitiveness in various sectors such as agriculture and food, among others.

The Institute approached Basetis in light of their need to speed up the process of classifying olive varieties for one of its projects. Selection is essential to maintain the Protected Designation of Origin of the olive oils produced. There wasn’t any programme to facilitate this work.

The solution

Our AI team developed a tool using artificial vision and deep learning capable of recognizing, classifying and segmenting the olives one by one on a conveyor belt.

To achieve this, a database of more than 26,000 images taken on site was built up, which made it possible to create an algorithm with 94% accuracy in classifying the olive variety. A special statistical application allowed the error to tend to 0.

This high degree of accuracy was essential to: secure the designation of origin of the olives and create an intelligent labelling that simplifies the process.

Results and products

5 people dedicated to the project 

  • Time reduction
  • Improvement of  the selection process
  • Process Automation
  • Certificate of Protected Designation of Origin

What was most innovative?

  • Creation of the first example of an olive sorter
  • Concatenation of a series of blocks (segmentation, classification, proportion estimation) to develop the solution
  • Creation of the database with different varieties of olives, as confirmed in the International Fair Expoliva 
  • Presentation of the results at an international scientific fair

What was most valued?

The partnership established with IRTA and the application of such cutting-edge technologies in the agri-food sector.

Technologies

  • Python 
  • Keras 
  • Tensorflow 
  • Image Segmentation
  • Edge detection
  • Data augmentation
  • Convolutional Neutral Networks
  • OpenCV

Services provided

  • Machine Learning
  • Statistical formulation
  • Deep Learning applied to Computer Vision