AI Is Transforming Quality Control in the Pharmaceutical Industry

Artificial intelligence improves the accuracy and efficiency of quality control in the pharmaceutical industry through MVTec’s HALCON software.

The Quality Control Challenge

In the pharmaceutical industry, quality control is an essential step in ensuring product safety and compliance with regulatory requirements. Even a tiny particle inside an ampoule containing a medicinal substance requires accurate detection and an appropriate response as part of the inspection process. One of the main challenges is distinguishing foreign particles from naturally occurring optical phenomena, such as air bubbles, reflections, and changes in lighting. In many cases, these phenomena have similar visual characteristics, and inspection requires an exceptionally high level of accuracy.

An industrial application documented by MVTec demonstrates how combining machine vision with artificial intelligence makes it possible to address this challenge and perform automated quality control of ampoules during the manufacturing process.

An AI-Based Solution

The system described in the application is based on MVTec’s HALCON software and incorporates deep learning technology, one of the advanced approaches in artificial intelligence. At the heart of the solution is semantic segmentation technology, which enables the system to analyze images at the pixel level and identify regions with different characteristics. By learning from examples, the system can distinguish foreign particles from bubbles, shadows, and reflections. This capability is particularly important when inspecting transparent products containing liquids, where defect detection is affected by the defect’s size, location, and the way light passes through the product. Using AI makes it possible to develop inspection systems capable of handling complex visual variations while maintaining consistency in the quality control process.

How Does the System Work?

The industrial application uses 12 industrial cameras connected to an industrial computer running HALCON software. Each ampoule is imaged up to 14 times from different angles to enable particle detection even when particles are located along the sides or at the bottom of the container. The images are processed and analyzed using a combination of traditional machine vision algorithms and deep learning models.

Artificial intelligence is used to detect particles and distinguish them from normal optical phenomena. In parallel, rule-based algorithms perform additional inspections, including detecting visual defects, checking the fill level, identifying color changes, and inspecting seal quality. Combining these technologies makes it possible to match each inspection task with the appropriate processing method while using AI capabilities for particularly complex tasks.

The Importance of Software and Expertise

The success of an AI-based application depends on the quality of the image processing software, the collection and preparation of training data, and the ability to adapt the algorithms to the product’s characteristics and the production environment. HALCON software provides a development environment for machine vision applications that combines deep learning capabilities with a wide range of tools for conventional image processing. In the application described, MVTec experts participated in preparing image datasets, classifying data, training the neural network, and evaluating model performance. The MVTec Deep Learning Tool was used to prepare training datasets containing images of defect-free ampoules alongside examples in which defects had been deliberately introduced under controlled conditions. Repeated training cycles made it possible to compare different models and improve the system’s ability to distinguish a defect-free product from one containing a defect. This case illustrates the importance of combining advanced machine vision software with professional expertise in designing, training, and implementing industrial AI solutions.

Application Results

According to the success story published by MVTec, the automated system significantly improved defect detection rates and reduced instances in which an existing defect went undetected. The inspection process benefited from improved speed, stability, and accuracy, alongside greater operational efficiency and improved product quality. This case demonstrates how AI can advance automation even in complex quality control tasks that require distinguishing between minute visual differences.

ASIO Vision’s Solutions

ASIO Vision markets MVTec’s software solutions in Israel, including HALCON, which combines advanced machine vision capabilities with AI-based technologies. These technologies are relevant to a wide range of industrial applications, including defect detection, quality control, product classification, object detection, and image analysis in automated manufacturing processes.

Through MVTec’s solutions, ASIO Vision enables manufacturers, system developers, and system integration companies in Israel to incorporate advanced AI capabilities into machine vision systems according to application needs and production environment requirements. The pharmaceutical industry application illustrates the technology’s potential and the possibilities it opens up for other industries seeking to improve the automation, accuracy, and efficiency of their quality control processes.

Source: A success story published on the MVTec Software website.

Facebook

LinkedIn

Technical PDF

PDF website

Website link

Share
Email
Facebook
WhatsApp
LinkedIn
Twitter
Print