New technology for manufacturing processes by Zebra

Zebra Technologies Corporation has introduced a new range of advanced AI features to its Aurora machine vision software, significantly boosting its deep learning capabilities for complex visual inspection tasks.

This enhancement is particularly relevant as, according to Zebra’s 2024 Manufacturing Vision Study, 61% of global manufacturing leaders anticipate AI will be a key driver of growth by 2029. Additionally, a Zebra report on AI usage in the automotive industry reveals that while AI and deep learning are already in use, there is a growing demand for these technologies to deliver even more advanced capabilities.

The updated Aurora software suite now offers powerful visual inspection tools designed to meet the needs of machine and line builders, engineers, programmers, and data scientists across various industries, including automotive, electronics, semiconductor, food and beverage, and packaging. It features a no-code deep learning optical character recognition (OCR) tool, drag-and-drop environments, and extensive libraries, enabling users to tackle complex challenges that traditional rules-based systems cannot easily address.

The Aurora Design Assistant, a part of the suite’s integrated development environment, allows users to create applications by arranging flowcharts rather than writing traditional code. The software also facilitates the development of web-based human-machine interfaces (HMIs) for these applications. The newest update introduces deep learning object detection and an enhanced Aurora Imaging Copilot companion application, which features a dedicated workspace for training deep learning models in object detection. Additional add-ons are available to support the training and deployment of deep learning models using NVIDIA and Intel GPU cards.

New advancements

Aurora Vision Studio offers a flexible, hardware-agnostic software platform that assists engineers in developing, integrating, and overseeing machine vision applications. The software’s user-friendly graphical interface allows the creation of complex vision applications without the need for coding, using over 3,000 pre-configured filters. The latest improvements include a new training engine designed to optimise results from low-quality datasets, along with support for Linux systems when performing inference tasks.

For developers working in C++, C#, and Python, the Aurora Imaging Library provides a comprehensive set of tools for processing and analysing 2D images and 3D data, employing both traditional methods and deep learning techniques. New additions to the library include anomaly detection tools designed for defect and assembly verification, using unsupervised deep learning. The OCR tool has also been upgraded to recognise a wider range of characters and symbols, leveraging a pre-trained deep neural network without the need for specific font instructions.

“Manufacturers across many industries face longstanding quality issues and new challenges with advances in materials and sectors such as automotive and electronics,” said Donato Montanari, vice president and general manager, Machine Vision, Zebra Technologies. “They are looking for new solutions that complement and expand their current toolbox with AI capabilities needed for more effective visual inspection, particularly in complex use cases.”