During the recycling process, one of the most important and least known tasks is the selection of packaging. The selection of containers is carried out in specific plants and its task consists of separating the different containers (from the yellow containers) according to their material and nature (usually plastics, metal containers and briks).

This work is of utmost importance since it is necessary to carry out recycling of these wastes in the proper way. In recent years, this process has evolved, going from being manual to being automated as Industry 4.0 technologies have improved and entered the boom, thus allowing greater efficiency throughout the process.

Technologies for packaging selection

During the selection of containers, different technologies are used that contribute in different ways to improving the efficiency and optimization of this process:

  • Computer Vision

    It consists of the application of various operations on previously acquired images in order to acquire information. Through this method, data such as the color of the observed object or its outline can be obtained. Using these techniques it is possible to determine the material from which the objects are made or to manipulate the image to extract the information that is most interesting for the application.

  • Machine Learning / Deep Learning

    Set of algorithms that are capable of learning a set of characteristics from which they can draw various conclusions, such as classifying or detecting objects. Of these techniques, neural networks stand out, which simulate the behavior of a human brain to learn from the data it receives. This method is very useful for the packaging selection process since it has a high percentage of success in this type of application, being essential to be able to separate waste and packaging in its corresponding category. It is commonly used in combination with Computer Vision when extracting information from previously processed images.

  • Robotics

    The use of collaborative robots is essential in the container selection process, since it allows automating the separation of containers in a physical way depending on the category to which they belong. In order to perform this function, it is usually used in combination with Computer Vision and Deep Learning, which provide the necessary information to locate packaging and waste, and the category to which they belong. They can also be used in combination with AGVs, in case it is necessary to transport the objects to another place.

Packaging selection implementation process

In order to establish a reliable and suitable selection method for use in the different plants, it is necessary to carry out a development in which data can be collected and the different technologies to be used can be configured. This process depends on multiple factors: the type of application to be carried out, place of installation, external factors … However, in almost all situations, specific steps to follow can be differentiated:

  • Data collection

    It consists of the acquisition of the information necessary to carry out the detection and classification necessary for the application. In the case of the selection of containers, it consists of taking images and data of the objects by means of cameras and sensors.

  • Labelled

    Method with which the characteristics that are interested in finding and to which class they belong are indicated in the acquired data. For a container selection application such as the one proposed, it is a question of indicating in an image where the objects we want to find are located and to what type they belong.

  • Training

    The tagged data is extracted by the detection algorithm used so that it learns the information to be detected. A properly trained system will be able to extract the characteristics that differentiate the different types of containers and will be able to locate and classify them for later separation.

  • Evaluation

    In order to confirm that the performance of the trained algorithm is correct, it is used with new data from which it is possible to obtain metrics that objectively qualify its performance. For the selection of containers, labeled images that have not been used in the training process will be used. In this way, it can be verified that it behaves in new situations by obtaining metrics such as precision or confusion matrices.

  • Installation

    Finally, once all of the above has been done, and the algorithm with the best evaluation has been chosen, the system is used in the real environment. For this, the algorithm is configured as necessary so that it returns the information that is interesting for the application and proceeds to use it accordingly. For the selection of containers, images of the waste will be taken, the algorithm will detect and classify them and send said information to the system, which will treat it as necessary to proceed with the separation of the corresponding waste.

At ATRIA we have developed a bulky waste identification system called SIARA that is based on artificial vision and Deep learning and that allows us to identify the objects that cause obstructions in current plastic waste separation plants. This process is important for the rest of the plant to function properly.

Benefits of using Industry 4.0 technology in packaging selection

As mentioned, the use of new technologies and automation are used to improve efficiency in packaging selection plants. This improvement translates into several benefits:

  • Possibility of recording data. As the entire process is automated, you can record the classifications made, the time they were made, or even store images that may be of interest.
  • Higher processing speed. As everything is automated, the system is able to perform the classification much faster than a person who had to do it manually.
  • By using vision systems, it is not necessary to employ large numbers of people for the task of identifying resdiues, regardless of the workload. So the staff can dedicate themselves to other functions.
  • Continuous improvement Being a learning-based system, changes and improvements can be made easily if necessary.

¿Necesitas ayuda?

¡Ponte en contacto con nosotros!