A machine vision insight from South Korea

Keunho Jeong, R&D team lead, at MVTec’s long-standing sales partner Teknix Co., Ltd. from Seoul, South Korea, is stating his point of view on machine vision and its impact on today’s industrial landscape.

Keunho Jeong R&D team lead at Teknix Co., Ltd.

COVID-19, which is not over yet, has made us radically change our living environment. The impact of COVID-19 in Korea began in 2020. Due to this, the economy has greatly contracted, and it has caused great changes in the industrial field. Representative examples were the food service industry and the travel industry. However, conversely, as non-face-to-face consumption of purchasing products on the internet increased, the logistics industry greatly expanded, and while the restaurant industry decreased, delivery services at restaurants increased.

Speaking of another change: from 2020 to the first half of 2022, office workers had to work from home and students had to study via distance learning. Meanwhile, sales of home appliances (laptops, PCs, TVs, etc.) increased, and due to eco-friendly policies, increased subsidies have caused a surge in demand for electric vehicles.

Machine vision in the packaging industry

Let us examine one more example from the machine vision sector. While the packaged ready-to-eat food industry expands, the demand for packaging inspection via machine vision increases. And this is where – for one customer – we came in with the following application: Many ready-to-eat products include plastic and plastic wrap. If these products are not properly packaged, food may deteriorate. It is not easy to inspect this with machine vision. In terms of 2D inspection, it was not easy because there are various printings for each product, and in 3D, there was a lot of noise data due to the non-flat characteristics of vinyl. And, there was a limitation that the production speed was slower than 2D image inspection. Therefore, 2D images were acquired by using a line-type laser. Using the obtained 2D image, a defect inspection was performed with deep learning. As a result, the existing defect detection rate has been greatly improved.

Machine vision of high importance for semiconductor and battery production

The machine vision industry in the Korean market has a very high portion of the semiconductor and display businesses. Due to this environment, investment plans in the memory/non-memory semiconductor industry began to increase in the semiconductor industry. However, while the display industry, which can be regarded as another main axis, has moved to the China market, investment in the display industry has decreased significantly. In the display industry without facility expansion, machine vision inspection equipment makers were shocked. So, they attempted to overcome this situation with a new business field. As displays and secondary batteries have very similar manufacturing processes, battery production for e-mobility became this new business field.

So, it was much easier for display manufacturers than for other equipment companies to enter the secondary battery inspection equipment market, where demand is greatly increasing.

The most sensitive parts of semiconductor and secondary battery inspection are primarily the speed and accuracy of defect detection. Of course, it is true that all machine vision is pursuing this. MVTec HALCON was optimal in these parts and showed greater power at larger image resolutions. Among the data structures of HALCON, there are “Region” and “XLD” types. Region has advantages in terms of speed, and XLD has advantages in terms of accuracy as a sub-pixel concept. With the combination of the two, it was possible to accurately detect foreign objects and measure objects at high speed. And, regarding the functional part as an example, HALCON provides a technology called photometric stereo. This enables data extraction of embossed or engraved data without using a 3D camera by combining 2D images obtained by using three or more lights at different locations.

This technology generally can be suitable for protrusion or dent inspection due to various pattern images in 2D images. In semiconductors and secondary batteries, this technology is widely used without using 3D cameras for chip and secondary battery surface inspection. It is also used for credit card OCR reading. And, additionally, deep learning is applied based on this data. The numerous features offered by HALCON are not limited to a specific application. Depending on where you use these features and how you combine them with various ideas, the results can be beyond your imagination.

Always keep an eye on new industry trends

In the past, semiconductors and displays were the main pillars of the Korea machine vision market, and based on this, it has expanded to the electric vehicle (secondary batteries, camera modules, etc.) and logistics markets. We need to concentrate on the MVTec’s machine vision software with its features is already ready to enter the robotics and embedded markets. There are functions such as Hand-eye calibration, which is closely related to robots, various matching techniques (2D/3D), 3D processing, deep learning, etc.

If we understand the trend of the times, always think about how to utilize the available machine vision technologies, and respond accordingly, we will be able to preoccupy new markets before anyone else.