The production line of any food industry performs work with different elements such as raw materials, semi-finished products and the final product. Each of these elements has certain characteristics, which make them more or less suitable for being handled by automated solutions. A prime example of this are agricultural or marine products, which, due to their heterogeneous and fragile nature, are often unable to be completely processed using automated solutions, and, if such solutions are available, these still tend to require considerable human intervention throughout the process.
Industry 4.0 has come up with solutions for the problems that arise when handling heterogeneous materials. The solutions proposed aim to automate an element of production, which, up until now, has been predominantly manual. These solutions look to combine different techniques, in particular the use of techniques such as “Machine Learning”, in order to allow for the advanced identification of the raw material/product.
The term, Machine Learning, simply speaking, refers to a series of new techniques in which a machine is “taught” to resolve problems that cannot be resolved by using classic techniques. This system is particularly useful when dealing with problems in which it is not possible to obtain good results by using mathematical tools, whether due to the large amount of data or to the specific characteristics of the products.
One of the fields in which the implementation of these techniques has really taken off is artificial vision. Artificial vision can be used to look for objects or to recognise complex patterns; a task which can prove problematic in productive processes which involve working with heterogeneous or un-positioned elements, for example, in the food or agriculture industries.
“Bin Picking” is a perfect example of how the aforementioned system can be used. The term “Bin Picking”, refers to the extraction of a specific element from a recipient that contains multiple items by using a robotized solution. This solution combines Artificial Vision tools with Machine learning techniques in order to find, position and determine the order of extraction from within the container. It also incorporates a robotic extraction system, which uses a claw that has been specially adapted for use with the element that it is going to be handling, and which is able to follow a specific trajectory depending on where the elements are located in the container. When working with raw materials, which are very heterogeneous in nature, searching for specific elements can be a highly complex process, given that the 3D vision techniques, which are generally used when searching for patterns, do not function properly with these types of products, therefore meaning that Machine Learning techniques must be used.
Thanks to the way in which these techniques are constantly evolving, in addition to the improved calculation capacity necessary for their implementation and the “democratization” of robotized solutions, the future for these types of solutions looks promising. By using these techniques, we can eliminate the need for operators to perform the tedious handling and supervision tasks that they are currently responsible for, and by combining these innovative techniques with the use of robotics, we will be able to improve the industry’s productive processes.