FOOD BOX SORTER

Real-time automatic image-based product classification system

Deep Learning is one of the branches of artificial intelligence that is increasingly present in automated manufacturing processes. Artificial vision based on Deep Learning offers a multitude of solutions to solve the most complex vision and classification challenges in a simple way.

This technique manages to analyse images in a very similar way to humans by training convolutional neural networks, an image segmentation system that quickly adapts to new samples without the need to reprogramme algorithms, just by re-training the system by introducing the images of these new categories to be added.

Among the most outstanding utilities of Deep Learning are:

  • Product classification and identification
  • Verification of the integrity of the product or packaging.
  • Locating, counting and checking the correct placement of features, parts and products in their assembly
  • Reading alphanumeric characters on complex surfaces such as transparent and soft bottles or labels with defects.

Therefore, machine vision based on Deep Learning offers solutions to problems that cannot be solved through traditional vision.

Previous
Next

In certain food processes in the meat sector where food is processed in line, it is common that at the end of the cutting lines it is the operators who identify the products and decide which further processing line they should go to.

This task is tedious and costly and can be replaced by a trained vision system for automatic sorting of the products, which allows the automation of the following processes. Food Box Sorter is an automatic product classification equipment in real time by Deep Learning. It is used to classify different product references or categories.  The equipment allows the classification process to be automated, reducing the error rate and significantly increasing effectiveness.

At the same time as classifying, the system detects quality defects, as once the product category has been identified, a specific quality inspection algorithm is executed, being able to detect contamination and defects in the slaughtering process.

This automation of the classification process in real time is achieved by training a neural network from thousands of images. Images are introduced and labelled indicating the reference with which we want the system to identify the product. In other words, the more images the system receives, the more this neural network is trained, until the system “learns” to identify the different product categories on its own.

What are its advantages?

Success rate over 95%

No reprogramming of algorithms necessary for introduction of new product categoriess

Reduction in labour costs

Functionalities

  • Product identification
  • Quality sorting
  • Detection of processing defects: bones, calves, cartilage..

Specifications

FOOD BOX SORTER