In our very first issue of Vision Spectra, published in the spring of 2019, our cover story delved into the role of AI and machine vision to perform defect recognition. Software vendors and system integrators had successfully integrated convolutional neural networks to label images to accurately classify defects. Industries such as food and beverage, automotive, and electronics were early adopters.
Nearly five years on, AI continues to evolve and mature.
The number of sample images needed to train a system has decreased
precipitously. AI is now used in perhaps unexpected ways, for instance as an adjunct to manual operations.
We revisit AI in the current issue as the theme of not one but two of our feature articles this month. Appropriately enough, one of them is authored by the same
company — MVTec — that penned our cover story in 2019. In this issue, “Vision Ensures Lithium-Ion Batteries Make the Grade,” author Klaus
Schrenker examines AI’s role in this increasingly important application, given the proliferation of electric cars.
AI is especially helpful for inspections at the back end of the battery assembly
process. When the cell stack is inserted into the cell housing and partly sealed,
different welds are needed to connect the current collectors inside the cells.
The quality of those connections is crucial to the quality of the battery, where even
a 1% fail rate is a serious issue. Deep learning is especially helpful in anomaly
detection — anomalies in the weld qualities — and for allowing previously unknown defects to be identified.
AI is also solving challenges in inspecting aluminum cans. One hundred billion aluminum beverage cans are produced every year in the U.S., and if dents or
defects occur during manufacturing, they can compromise the quality of the
beverage. Traditional rule-based inspection is painstaking to set up and maintain, given the parameters that affect the required sensitivity and accuracy of vision
systems. Predictive AI and deep learning can be used to recognize defect types,
with AI models created by capturing image samples and using these images to
train models.
Only a few years ago, skilled programmers with significant data analysis were needed to build a model to effectively recognize and label a defect. Today, given
the rapid growth in off-the-shelf software for nontechnical users, models can be
created without programming or coding. Read more here in “Detecting
Dents and Damage in Aluminum Cans Using AI Computer Vision,” from Teledyne DALSA’s Szymon Chawarski.
Enjoy the issue!
