summary-header

Automated quality control of animal feed production

Computer vision service developed by Rubius helps determine the size and shape of pet food pieces 6 times faster

Technologies

  • C#
  • Python
  • .Net
  • React

Customer

The customer is one of the largest food manufacturers worldwide.

Problem

The customer found out pets prefer food of balanced-size pieces. And if there are too many large pieces in the package, pets may not finish their meal. That is why pet owners could choose products from other manufacturers. It can lead to an outflow of customers and negatively affect the company’s profits. 

Previously, company employees checked the size of pet food pieces manually during the quality control stage. Every three hours, a quality control specialist would take a small part of the batch from the continuous conveyor belt and separate the longest pieces. If the number of such pieces exceeded the acceptable limit, it indicated that the cutting mechanism was not working properly.

For the next batch, the employee adjusted the parameters of the cutting mechanism and the selected pet food was sent back for processing. The checking process needed acceleration and improvement to ensure that every package has feed of required size in optimal quantity.

We have identified three main factors that cause large pieces of pet food entering the packaging:

  • The subjectivity of assessment: the quality control of the same piece of pet food can vary due to qualifications and experience of the employees. It makes the process inconsistent.
  • Non-reducibility of the system: a person cannot perform the same task with the same accuracy for a long time. Human factors, concentration problems, and tiredness lead to inaccuracies in quality control measurements.
  • Low control speed: It takes 15 to 30 minutes for a quality control specialist to test the batch.

Solution 

The Rubius team developed a special computer vision (CV) module for quality control of feed. The module is integrated into a web-based application which helps the specialist to track the size of food pieces based on photos from cameras located along the conveyor belt. The application can be launched on the tablet and PC.

How it works

  • The user is authorized through a common portal of the producer’s site ecosystem. The application downloads the necessary manufacturing data where the user is working. All measurement data are sent to this factory’s database.
  • Then the user selects a work shift, sets the initial parameters and corrects the source data. Now everything is ready to start work.
  • The cameras along the conveyor belt photograph the feed by pressing the "Take a photo" button
Screen for Initiating Photo Processing with Computer Vision
Photo data processing start screen
  • The system displays the result in the form of an image with markings on the pieces of pet food, as well as a text message indicating whether the part of the batch tested meets the benchmarks.
Result of Inspecting the Control Section of a Batch Using CV
The quality checking of the batch with computer vision: the results screen

The results screen displays the following data:

  • A, B, C…G – percent of pet food pieces meeting specific length or width requirements as specified.
  • High (red light) – the allowed value is exceeded by the percent of pet food pieces with this size and shape.
  • OK (green light) – the percent of pet food pieces with this size and shape is in accordance with the norm.
  • Low (yellow light) –  the percent of pet food pieces with this size and shape is too low.

For a more detailed check, the user can view the summary table with data on the specification (length, shape, area) of the pet food pieces and compare them with the acceptable values. It helps to quickly correct irregularities in the cutting mechanism and ensures that the pieces of food about the required size enter the package. The total time for verification takes less than 5 minutes.

The web application has a help feature that displays detailed instructions on how to use the application and how to evaluate pet food pieces. It also contains data on past measurements.

Result

The CV-based web-application helps quality control specialists track the condition of the pet food at any moment of conveyor belt movement. It helps quickly identify the size of the food pieces and correct the irregularities of the cutting mechanism.

The assessment of the feed quality occurs according to the permissible indicators (length, shape, area), they do not change over time. This process gives regular and accurate results.

The system operates around the clock with the same productivity in all factories of the pet food manufacturer. This allows to increase the speed of the feed quality control by 6 times. In this way, optimal-sized pet food pieces enter the package with high accuracy. It helps to increase customer loyalty and reduces the outflow.

Let’s discuss your project

Tell us about your requirements and we’ll get back with a possible technical solution

Required field


Call us

Write to us