ANALYSIS: DEEP LEARNING
How can deep learning optimise vision inspection?
Hongsuk Lee, founder and CEO of Neurocle, explores the many benefits of deep learning, as well as the roadblocks preventing its widespread adoption in machine vision
T
he manufacturing industry is constantly looking for new ways to improve the quality and efficiency of
its products and processes while reducing costs. One of the emerging technologies that can help achieve these goals is deep- learning-based vision inspection. Vision inspection is a crucial process
for ensuring the quality and performance of various products, especially in sectors such as automotive and battery industries, where safety and durability are paramount. However, traditional machine vision systems, which rely on predefined rules and algorithms, often face limitations and challenges when dealing with complex and variable scenarios. For example, inspecting the surface finish, welds, soldering points, glue, coating, and other amorphous shapes of automotive and battery components can be difficult for rule-based systems, as they may not capture all the possible defects and variations. Moreover, traditional machine vision systems may not be able to adapt to changing environmental conditions, product specifications, and customer demands. The global machine vision market size is expected to expand at a compound annual growth rate (CAGR) of 12.3% from 2023 to 2030, having been valued at $16.89bn in 2022. This growth is now being driven by the advancement of AI/deep learning technologies and the emergence of new applications and markets, as well as the increasing demand for quality inspection and automation. In this article, I will explore some
of the latest applications and benefits of deep learning for vision inspection, offering examples from the automotive and battery industries. I will also discuss some of the roadblocks and solutions for its widespread adoption, and share some experience and insights as the
‘Deep learning can provide more accurate inspection results, capturing subtle defects that may be missed by traditional vision systems’
head of a company providing AI deep learning vision software across various manufacturing sectors.
Benefits of deep learning in vision inspection Improved quality and performance Deep learning can provide more accurate and consistent inspection results, as it can capture subtle and complex defects that may be missed by human or traditional machine vision inspection. This can lead to improved quality and performance of the products, as well as increased customer satisfaction and loyalty.
Reduced costs Deep learning can also reduce the costs associated with vision inspection, as it can automate and streamline the inspection processes, reducing the need for human intervention and supervision while increasing inspection speed and efficiency. This reduces the costs of rework as it can detect and prevent defects early in the production process and ensure that only high-quality products are delivered to the customers.
Increased flexibility and scalability Deep-learning-based vision systems can adapt and learn from new data and handle changing environmental conditions.
10 IMAGING AND MACHINE VISION EUROPE DECEMBER 2023/JANUARY 2024
They can also handle different types of products, materials, processes, and support different models and variants of the products. The inherent adaptability of deep learning technology offers a distinct advantage in terms of increased flexibility and scalability. Once implemented, machine builders can seamlessly extend these models across all their equipment without the need for extensive reprogramming or system reconfiguration. This scalability enables the incorporation of the deep learning model into various equipment, empowering manufacturers to effortlessly apply the technology to diverse products, materials, and processes. Moreover, this adaptability allows the system to learn from new data and effectively navigate changing environmental conditions, providing a future-proof solution for evolving industrial requirements.
Examples of deep-learning- based vision inspection in the automotive industry One of the main applications of deep learning for vision inspection in the automotive industry is body-in-white inspection. This is the stage where the metal frame of the car is assembled, and it involves a large number of welds that need to be inspected for quality and consistency. Deep learning can improve the detection of defects such as cracks, holes, gaps, spatter, and misalignment in the welds, as well as angles of the weld seams. The significance of deep learning
extends beyond the assembly of the car frame to encompass the critical stages of paint application and final assembly. Achieving a high-quality surface finish on the car body, vital for both aesthetic
@imveurope |
www.imveurope.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32
