Introduce tu búsqueda a continuación y presiona el botón para obtener resultados.
One of its lines of work is the development, production and commercialization of in vitro diagnostic tests for the food and environmental sectors.
ZEULAB is a Spanish biotechnology company specialized in the development of effective solutions that help its customers in the control of food safety through the application of the latest technologies. One of its lines of work is the development, production and commercialization of in vitro diagnostic tests for the food and environmental sectors.
In the manufacturing process of some of these products, different quality failures occur in their products at a functional level and/or at an aesthetic level with different casuistry in terms of product wastage.
These quality failures could have a significant impact on the company, either in terms of brand perception or product reliability. In addition to an economic impact on their costs by having to withdraw these products either from production or from the market, generating significant logistics and replacement costs.
This is why ZEULAB takes great care in this aspect by carrying out different quality processes to detect and, if necessary, recall all defective products.
C/Bari, 25 Duplicado 50197 Zaragoza, Spain
The objective of this project was to work on the automation of the detection of these quality failures, validating if through the use of artificial vision technology, combining classical and Deep Learning techniques, it was feasible to eliminate or reduce as much as possible these quality defects in their process.
In order to evaluate the algorithms, it is necessary to first carry out a data acquisition process, i.e. to build a bank of labeled images (dataset) with which to evaluate the most suitable algorithms. This process was carried out by means of an image capture system with high resolution cameras.
Subsequently, different classic algorithms used in computer vision and other deep-learning algorithms were developed and tested for accuracy using this image bank.
The result obtained in the project allowed, on the one hand, to find the types of quality failures that are more feasible to detect with vision and artificial intelligence and which are more complex. With the final algorithm developed, it was possible to perform a preliminary detection of the highest priority faults in order to find a subsequent industrializable solution.
