The Spanish company Robotics & Vision Technologies (RVT) is a partner company of our distributor INFAIMON S.L. and one of MVTec’s Certified Integration Partners. RVT develops applications that enable the improvement of automated processes within the industrial sector by using artificial vision, robotics, and augmented reality. For a leading manufacturer of frozen dough and bakery products in Spain, they developed a quality control system based on machine vision and deep learning algorithms.
The biggest challenge of the project was to achieve the highest level of quality control in 100% of the customer's production. For this purpose, it was necessary to control all visual parameters of the product in different phases of production. This implies metrology parameters (2D and 3D), analysis by vision of measurable parameters, and non-measurable parameters analyzed with artificial intelligence. In addition, there were many difficulties regarding the identification and correct classification of the products, which can be attributed to several causes:
- Natural and handmade products, all different from each other
- Metrology control of various parameters depending on the type of product
- Control of subjective parameters such as the degree of toasting.
Tasks of the system
The developed system performs the following tasks:
- Identification of the product in the production line.
- Carrying out various measurements on each product depending on the product type.
- Classification of the product according to different visual parameters using artificial intelligence, i.e., deep learning.
This system is able to analyze up to 140 pieces per minute 24 hours a day. It automatically sorts out the non-conforming products. Practically no maintenance is required.
The software was developed with substantial use of the MVTec HALCON libraries. RVT has developed its own libraries to create an interface tailored to the customer's needs and a wizard that makes it easy to add new references. Within this solution, HALCON performs 2D and 3D metrology as well as deep-learning-based segmentation and classification. So, the segmented breads (using traditional machine vision approaches) are classified according to their visual appearance. Without the integrated deep learning functions this would not have been possible or would have involved a very high programming effort.
The complete system consists of high-resolution color line scan cameras and specific lighting that allows the analysis of very subtle color and texture parameters. It also includes 3D metrology systems and high-performance processing units.
Advantages of the solution
- Control of all quality parameters
- Released capacities can now be used elsewhere.
- Increase in production by 30%
- Analysis of 100% of production
- Implementation of metrology controls that could not be carried out previously.
Text and images are kindly provided by Robotics & Vision Technologies.