How Machine Vision Is Changing the Textile Industry

AI stands for Artificial Intelligence. It refers to a computer algorithm that automatically identifies objects/pattern in digital data, for instance, a digital image, and then classifies them according to learned criteria. Based on the classification it initiates and carries out tasks.

As an atmospheric scientist, I have been coding such computer programs to detect air pollution changes over time. In my MS thesis, I developed an automatic detection algorithm to classify NOAA satellite data over the Arctic  into sea-ice, open ocean, land, water clouds, mixed phase clouds and ice clouds. Today, the textile industry makes use of AI and machine vision at various steps in the production and even retailing process. Read to learn how AI helps the fashion industry to reduce costs and waste.

Disclosure: Sponsored post.

What Is Machine Vision and AI?

Your digital camera or the scanner on you phone are basically a machine that takes visual data. When you would feed these data to a computer, a well-trained AI software could identify what the machine “saw”. Depending on the task the programmer chooses between supervised and unsupervised learning they can create AI for various purposes. In principle, three different mathematical methods exist to teach the computer to distinguish different features and categorize them:

• Statistical methods
• Fourier-/wavelet analysis, and
• Application of models

Which of these methods is the most suitable depends on the application for which the AI is designed. While these methods differ strongly from a mathematical point of view, they have the following in common. The computer “learns” how certain things look like in the digital data. When the data differ from the first learned feature, the AI can store where the deviation occurred, and checks whether the data match another learned feature, and so on. Finally, the software may even take pre-coded actions depending on the results.

Note that AI is also occasionally referred to as machine learning or deep learning.

How Machine Vision Is Used in the Textile Industry

Depending on the device, machine vision data are taken line-by-line or as an area of X pixels times Y pixels. In the textile industry, application the latter is more common.

Quality Assurance/Quality Control of Yarn

Spinning creates yarn. During the spinning process, short fibers may lead to uneven structure of the yarn resulting in different hairiness. Also uneven twisting might occur which leads to to thin or too thick places or neps. These defects would lower the quality of the knitted or woven fabric, and finally the garment. Teaching the computer how the high quality yarn should look like in the digital image serves to identify yarn of minor quality before it gets on the loom. Additional learning permits even the classification of the kind of defect.

Inspection of Fabric Quality

Since the Industrial Revolution in the 18^th century, machine looms weave the fabrics. Because of this automatization, human workers have to inspect the fabrics for quality. While the fabric passes at a given speed in front of these works, they have record defects like loose weft yarn, yarn pulled with the shuttle carrying pirn in the wrap direction, oil spots from the machinery, ripped yarns or fabric areas.

Their detection rate amounts 60 to 75% on average due to fatigue, the speed of the moving fabric (10 m per minute), and the boring task. On the contrary, well-trained AI software has detection rates up to 98%.

Machine Vision and AI for Trend and Color Detection on the Runways

Typically fashion shows are a firework of colors and styles. While the human eye can recognize colors, it is hard to grasp the exact wavelength of the color and which colors occur the most. The background also affects the human perception of color as the illusion of the blue-black and white-gold dress demonstrated about a decade ago.

However, AI can remove the background from the machine vision data, filter the light, and objectively recognize the colors including providing the percentage of the colors in the image.

Some fashion brands (e.g., H&M, Zara) as well as designers, retailers, producers use AI tools to identify changing and new trends. As a result, they can respond to these changes faster than their competitors. Consequently, they cannot only sell more products, but also save because they can stop the production when the trend slows down. The environment also benefits because the AI technique reduces the amount of overstock that becomes waste.

AI and Machine Vision Enhance Customer Experience

Machine vision combined with AI serves to give online customer a virtual in-store experience. The AI identifies the style and type of the garment you are looking at. Based on that knowledge, the AI searches for similar items in the online store to recommend them to the customer. You probably remember having seen the sentence

Customers who bought this item also bought XYZ.

The Scanner Combined with AI as an Accountant for Inventory and Restocking

At checkout, the scanner not only reads the price, but also provides input data to the AI software to track inventory, initiate restocking, and provide insight on how popular a garment is by determining the time between stocking and selling.

Identification of Knock Offs

Obviously, high-end brands suffer from duplications around their products. These low quality knock-offs can jeopardize the customers’ faith in the brand and the brand image. Here deploying machine vision and AI can help in the identification of fake products, and logos. Consequently, the brand can remove the fake products before they hit the market.

Advantages of Machine Vision and AI in the Textile Industry in a Nutshell

Early detection of defects reduces the costs, waste, and delays in production time. The textile industry can recycle waste in the spinning process, while woven fabrics require transport to enter the recycling chain as filling material, for instance. Furthermore, automation of mundane, repetitive visual inspection increases the detection rate of defects, reduces the costs, and is objective. However, application of machine vision with deep learning in the inspection process requires different software depending on the fabric type, weave, pattern, and print. Consequently, the upfront costs for creating the AI may be steep.

References

Azim, G. (2015) Identification of Textile Defects Based on GLCM and Neural Networks. Journal of Computer and Communications, 3, 1-8. doi: 10.4236/jcc.2015.312001.
Lai, P., Westland, S. (2020) Machine Learning for Colour Palette Extraction from Fashion Runway Images, International Journal of Fashion Design, Technology and Education, 13:3, 334-340, DOI: 10.1080/17543266.2020.1799080
Langs, G., Röhrich, S., Hofmanninger, J. et al. (2018) Machine Learning: From Radiomics to Discovery and Routine. Radiologe 58, 1–6. DOI: 10.1007/s00117-018-0407-3
Pereira, F., Carvalho, V., Soares, F., Vasconcelos, R., Machado, J. (2017) Computer Vision Techniques for Detection of Yarn Defects. In: Wong, W.K., Applications of Computer Vision in Fashion and Textiles, Woodhead Publishing
Rahman, M.M. (2021) Applications of the Digital Technologies in Textile and Fashion Manufacturing Industry. Technium: Romanian Journal of Applied Sciences and Technology, 3(1), 114-127. Available at SSRN
Wong, W.K., Jiang, J.L. (2017) Fundamentals of Common Computer Vision Techniques for Textile Quality Control. In: Wong, W.K., Applications of Computer Vision in Fashion and Textiles, Woodhead Publishing
Wong, W.K., Jiang, J.L. (2017) Computer Vision Techniques for Detecting Fabric Defects. In: Wong, W.K., Applications of Computer Vision in Fashion and Textiles, Woodhead Publishing

© 2013-2023 Nicole Mölders | All rights reserved